William Sayers

2009
Milestone 3

05025397

Genetic Algorithms and Neural Networks

Genetic Algorithms and Neural Networks
William Sayers (05025397)
Supervised by: Colin W. Morris (BSc, MSc)
Backpropagation neural networks are usually trained using some iterative method derived from a mathematical analysis of the operation of the network. This technique is time consuming and requires understanding of the mathematical principles involved. This project will investigate the application of genetic algorithms to replace the "normal" training of the network. This involves setting up some candidate networks and then picking the best of these. The best of the population are then combined in some way to produce a new population of candidate networks. This procedure is continued until a satisfactory network is obtained.

William Keith Paul Sayers225 Page 1 of Faculty of Advanced Technology 22/04/2009

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3

Contents
Contents ............................................................................................................................ 2 1 - Introduction ............................................................................................................... 10 2 - Research .................................................................................................................... 10 2.1 - Neural Networks ................................................................................................. 10 2.1.1 - The Biological Neuron................................................................................. 10 Figure 1 - A human neuron ............................................................................................ 10 2.1.2 - The Artificial Neuron................................................................................... 11 Figure 2 - An artificial neuron ........................................................................................ 12 Equation 1 - Summation of the inputs and weights ........................................................ 12 Figure 3 - The sigmoid function shape ........................................................................... 13 2.2 - The History of Neural Networks ........................................................................ 13 2.3 - Advantages and Disadvantages of Neural Networks ......................................... 14 2.3.1 - Advantages................................................................................................... 14 2.3.2 - Disadvantages .............................................................................................. 15 2.4 - Current Applications of Neural Networks .......................................................... 17 2.4.1 - Neural Networks in Medicine ...................................................................... 17 2.4.2 - Neural Networks in Business ....................................................................... 18 2.4.3 - Object Trajectories....................................................................................... 19 2.4.4 - Robot Control .............................................................................................. 19 2.5 - Back-propagation ............................................................................................... 20 2.5.1 - Back-propagation overview ......................................................................... 20 Page 2 of 225

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3 2.5.2 - Back-propagation in depth ........................................................................... 20 2.5.3 - Back-propagation library ............................................................................. 22 2.6 - The FANN neural network library (Nissen, FANN) .......................................... 22 2.7 - Genetic Algorithms ............................................................................................ 22 2.7.1 - History of Genetic Algorithms .................................................................... 23 2.7.2 - Advantages and Disadvantages of Genetic Algorithms .............................. 24 Figure 4 - local optimum (units are undefined in this case). .......................................... 27 2.8 - Training Neural Networks with Genetic Algorithms ......................................... 28 2.8.1 - Determining Weight Values with Genetic Algorithms................................ 28 2.8.2 - Representation of a Neural Network within a Genetic Algrithm ................ 28 2.8.3 - Using Genetic Algorithms to Determine Neural Network Structure ........... 29 Figure 5 - A feed-forward neural network ..................................................................... 30 2.9 - Cascade Correlation............................................................................................ 32 2.10 - C# User Interface Programming ....................................................................... 37 2.11 - Writing C++/CLI Wrappers for Unmanaged C++ code................................... 37 2.12 - Application of Research ................................................................................... 38 3 - Design ....................................................................................................................... 38 3.1 - Program Requirements ....................................................................................... 39 3.2 - Design of the Class Structure ............................................................................. 39 3.3 - Linking C# code to managed dll’s...................................................................... 40 3.4 - Design of the User Interface ............................................................................... 40 Page 3 of 225

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3 3.4.1 - Main Form ................................................................................................... 40 Figure 9 – Network and output tab design (Main form) ................................................ 41 Figure 10 - Mean squared error tab design (Main form) ................................................ 42 Figure 11 - Dataset tab design (Main form) ................................................................... 42 3.4.2 - New Network Wizard .................................................................................. 43 Figure 12 - Design of the training algorithm tab (Wizard form) .................................... 43 Figure 13 - Design of the network settings tab (Wizard form) ...................................... 44 Figure 14 - Dataset selection tab design (Wizard form)................................................. 44 3.4.3 - Working Form ............................................................................................. 44 Figure 15 - Working form design (Working form) ........................................................ 45 3.4.4 - About Form .................................................................................................. 45 Figure 16 - About form design (About form) ................................................................. 45 Figure 17 - Dataset display design (About form) ........................................................... 46 Figure 18 - Licenses display design (About form) ......................................................... 46 4 - Implementation ......................................................................................................... 46 4.1 - User Interface Implementation ........................................................................... 47 4.1.1 - Main Form ................................................................................................... 47 Figure 19 - Network and Output tab............................................................................... 47 Figure 20 - Mean squared error graph tab ..................................................................... 48 Figure 21 - The dataset tab ............................................................................................. 49 4.1.2 - New Neural Network Wizard ...................................................................... 50 Figure 22 - Learning Algorithm tab ............................................................................... 50 Figure 23 - Network Settings tab.................................................................................... 51

Page 4 of 225

.........1...... 57 4...About form.............Testing Data .......................... 54 4.............................1......................................................... 51 4. 53 Figure 27 .....Genetic Algorithms ...................... 52 Figure 26 ..................... 54 4........................................................1......................................................About form...........Back-propagation ................ 61 Figure 31 ........................1..........................Dataset tab ...................The mean squared error at testing (XOR).......................................2 ..................6 ..... 62 Page 5 of 225 .......................Back-propagation and Cascade Correlation Implementation ................Mean squared error upon last training test (XOR) .................................................................... main view ......... 60 5........ 56 4..4 .About form........................................................................................ 59 5...Working Form ....................................................................................................Cascade Correlation ..................4 ............. Licenses view ..About Form ................................2 ................... 54 Figure 28 ..................1 ..............................5 ..............................................Passing Information between Forms .......1 .................William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Figure 24 ................................................................ 54 4....................Working form ......2..............................................1....................XOR .....1.................Data Parser Implementation ......................................................... 52 Figure 25 ........................ 61 Figure 30 ..............Keeping the User Interface Active whilst Processing is Occurring..................3 ..................... 58 5..... 57 5 ................4 .... 59 5............Graphs ...............................Neural Network Implementation ............................ 61 Figure 29 ........................1............ 55 4.................. 58 5....................1 ..3 ................................3 ..............1...................................... Datasets view .............................................................................................................. 56 4...........................Number of iterations to achieve target (XOR) ...........Genetic Algorithm Implementation ..

........... 69 6............................................. 64 Figure 33.............................................................................Back-propagation and Genetic Algorithms ......................Number of iterations to achieve target (Viruses) ......Virus Classification .............................Fishers Iris data.......1 ..............Cascade Correlation and Genetic Algorithms ...Genetic Algorithms ...........1 ..........................................................3 .............................................................................................. 69 6............ 67 Figure 36 ................3..............1.......................3......................................William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 5.....................................2 .....................Number of iterations to achieve target (Iris data) ................................. 65 5..........................................2 .Fishers Iris Data ................. 63 5.........................Mean squared error upon last training test (Viruses)..........4 .............................................................Back-propagation ...2 ......1 .........................The mean squared error at testing (Iris data) .................Graphs .......... 63 5..................................... 67 Figure 35 ........................... 66 5.. 69 Page 6 of 225 ........ 68 6 ....Graphs ..............4 . 67 Figure 37 .........................................................Virus Classification.2.....................................................................................................................Back-propagation .............................Genetic Algorithms .....Mean squared error upon last training test (Iris data) ................... 64 Figure 32 ............1 ....1...............................1...........................................................................2 ............Comparisons ................................2......2. 66 5..............................Cascade Correlation ...............................2 ..................3..XOR ...................3 ....3.......3 ...................... 63 5.............. 68 6......................................................................2..... 64 Figure 34................ 63 5................................. 65 5.............................Mean squared error at testing (Viruses)........................................................3 ...... 66 5..........................Cascade Correlation ............... 68 6..................... 68 6........................

...3 ................................................. 174 9....................................................Model .............................................Bibliography............................... 73 8.............................2... 207 9.................Libraries Researched .........................3 .......Virus Classification....................Objectives .............................. 74 9......... 209 9.......... 71 8 .............. 109 9....................................Possible User Interface Improvements .....1......................................1 ....................... 204 9.........Appendices .................... 70 6..Testing conclusions ...............................................................................Evaluation ........................................................................Datasets...............1 ............................................................... 73 9 ................. 72 8..........................................................How many of the original objectives were achieved? ..... 75 9.................................................... 210 10 ..................................................4 ................................UML Class Diagram... 69 6...............................................................................2 ..........1 .............................................................1 ...................................................1..........3 .........2........................................................................................... 69 6.............2......................................................................Works Cited ...................................................... 158 9..................................................Possible Improvements .. 214 Page 7 of 225 ........Source code ........5 .............1................................................................................................................................Possible Training Algorithm Improvements ........... 70 7 .............................William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 6........View .........2 ................................... 211 11 .......3 ......XOR ...................................................Controller .................... 74 9.......DLL Wrapper for FANN functions ..........................2 ..2 ..........................................................................................................................4 ......................... 70 7.Fishers Iris Data .......1..1 .............................

............................................................ 219 12..............................................2 ........................................................................ 220 13 ...............................................3 ................. 217 12...................Tables ......Table of tables ................................... 217 12............................1 ........ 222 Page 8 of 225 ........Table of Figures ......................................................................Diary......................................................................Table of Equations ..............................William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 12 .......

.................... Student Name (Printed) Student Signature Registered Scheme of Study Date of Signing ………………………………... and that neither this project..................... Any material taken from published texts or computerised sources have been fully referenced.................. Page 9 of 225 .. ………………………………. nor any part of it.... the work described within this project is the result of the investigation carried out by myself........William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 STATEMENT OF ORIGINALITY SCHOOL OF COMPUTING DEGREE SCHEME IN COMPUTING LEVEL THREE PROJECT This is to certify that.... ………………………………........ and I fully realise the consequences of plagiarising any of these sources...... except where specific reference is made............. ………………………………....... has been submitted in candidature for any other award other than this being presently studied..............................

The Biological Neuron The human brain is composed of many millions of nerve cells called “Neurons”. In addition.A human neuron Page 10 of 225 . the report will investigate. but in combination with a network of other neurons within the brain.1 . compare and contrast three common learning algorithms that are applicable to neural networks.Neural Networks Neural networks are networks of small-specialised units that sum the products of their inputs and the weights on those inputs. This report encompasses design. Although they are simple in concept. Figure 1 . they can produce immensely complex results.1. development and final prototyping of the program developed for this project.Research 2.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 1 . on the course “Computer Games Development”. 2 . 2. Neurons do not make decisions alone.1 .Introduction The purpose of this report is to bring to a conclusion my final year project on neural networks and genetic algorithms.

the nucleus will generate an “action potential”. an activation function and a method for representing an output – either to another artificial neuron in the next layer.1. Axon branches can connect to other types of cells also. as mentioned above).2 . The axon and dendrites both split before terminating into “Axon branches” and “Synapses”. with extensions called “Dendrites” which are the neuronal inputs and a single extension called an “Axon” which is the output for that neuron. or to another part of the program. (Barber.The Artificial Neuron In order to construct an artificial neuron we need a way of representing the same model in data structures and algorithms. The usual method for approaching this is to have a number of inputs. a number of weights associated with those inputs. AI: Neural Network for Beginners pt. a synaptic gap occurs between the axon branch and the synapse. 2004). (Bourg & Seeman. a summing function. Axon branches connect to synapses to form the neural networks within the brain. If the inputs from the dendrites to the neuron nucleus when summed are past a certain threshold. 1. Page 11 of 225 . In the case of the axon branches connecting to synapses that are part of a second neuron. 2006) 2.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 A neuron is composed of a cell body. Concerning inputs from the dendrites. in this gap between neurons chemical reactions take place that will inhibit or excite the signal. This action potential extends down the axon to the cells at the end of the axon branches (often synapses connecting to other neuronal cells. the nucleus of the neuron cell processes them and produces an output. in order to allow the brain to control cells external to the brain.

Summation of the inputs and weights The next stage depends on the activation function chosen. A stepping function can be used which will output zero (or non-activated) if the summed products are below a certain value. Alternatively. A specific input accepts a data item (usually a binary digit.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Figure 2 . The activation function accepts this stored value. Page 12 of 225 . ℎ = × ℎ Equation 1 . or integer). real number. a differentiable function of some kind (such as a sigmoid function) which will output values between zero and one depending on the value of the inputs is another possibility. or one (activated) if the summed products are above a certain value. The product of the data item and its associated weighting is then stored before the summation of all the data items and associated weights occurs.An artificial neuron The system devised with these components is as follows.

2 1 0.8 0.4 0.2 .A sigmoid function (Bourg & Seeman. in 1958. however he was unable to come up with a reliable mathematically accurate mechanism for allowing multilayer perceptrons to learn (Rosenblatt. In this example “y” is the function output and “t” is the function input. contributed to the ideas that underpinned early neural network research.2 0 -6 -4 -2 0 2 4 6 Figure 3 . 2006) 2.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 1. the work of 19th century psychologists such as Freud. 2004).The History of Neural Networks The foundation of neural network research are in psychology. 1. 1958). (Bourg & Seeman. (Barber.6 0. 1943). Frank Rosenblatt developed the perceptron model then. AI: Neural Network for Beginners pt. 2004).The sigmoid function shape = 1 1+ Equation 2 . Page 13 of 225 . William James and others. as defined in the graph above. McCulloch and Pitts formulated the first neural network model which featured digital neurons but the neurons had no capability to learn (McCulloch & Pitts.

3 .3. 2. Parker discovered it independently in 1982 whilst performing graduate work at Stanford University in the United States of America and called it “learning logic”. taking Page 14 of 225 . Back-propagation allows for the training of multilayer perceptrons and in particular. (Blum. perceptrons with “n” hidden layers.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Two layer perceptrons were the subject of several experiments to determine their usage and effectiveness. The individual neurons in a neural network have no time dependencies on each other and each can therefore be run on a separate processor (or as separate threads on a single processor) if desired without causing the problems often associated with such parallelism.1 . 1982)). 1992). The next major advancement in neural networks was not until 1974. A good example of this is that neural networks have been implemented to run on architecture originally designed for processing three dimensional computer graphics. 1974) discovered the back-propagation algorithm (independently rediscovered by Parker in 1982 (Parker. Werbos discovered the algorithm whilst working on his doctoral thesis in statistics and called it “dynamic feedback”. of “n” nodes.Advantages Neural networks make possible or practical many things that with conventional solutions would be extremely difficult. One of the benefits of neural networks when compared with other problem solving techniques in that they are inherently parallel and thus can run very effectively and efficiently on parallel hardware. where “n” is an undefined number.Advantages and Disadvantages of Neural Networks 2. used on this architecture they become an extremely fast as well as flexible tool. when Werbos (Werbos.

William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 advantage of the massively parallel architecture of this hardware and the programmable sections of the graphics pipeline (vector and fragment shaders) to produce extremely fast neural networks (Davis. We can simply present the neural network with all the data as opposed to following the circuitous and overly complex (by comparison) statistical approach demonstrated in “Table 1 . 2001). 2. where explaining why you have arrived at a particular diagnosis is an important part of the process.Disadvantages One disadvantage to neural networks can be that it is very difficult for a human being to analyse a trained neural network and describe why it may come up with one answer over another. 2 Formulate a statistical model. 4 Analyze the results. 1 Decide on relevant data. This is a large disadvantage in areas such as medical diagnosis. When using a neural network there's no need to establish before attempting problem solving which data is relevant. in fact. Neural networks can excel at determining what data in a particular set is relevant.The statistical approach to deciding relevant data”.The statistical approach to deciding relevant data Another advantage of neural networks is that noisy data is not a real problem for them to learn to interpret. 5 Build a system that incorporates what we have learned. Table 1 .2 .3. to attempt to provide a more robust post-training network. noisy data is utilised during training upon occasion. Page 15 of 225 . as irrelevant data will simply end up weighted so that it has zero or close to zero actual effect on the solution produced by the network. 3 Run the formulated model.

a sunny day. they tested the network with the remaining fifty images that the neural network had not seen before and the neural network correctly identified most of the tanks. asked to differentiate between the two sets of pictures. it can also be hard to tell how well (or how badly) trained a neural network is. When using neural networks is that. Therefore. the network had trained itself to recognise whether the sky was cloudy or not. The training times on neural networks can also be a disadvantage. Page 16 of 225 . Eventually the original data set was re-examined. although the length of time the problem would take to solve without employing a neural network must be taken into account when deciding how much of a disadvantage. there is no formal methodology with which to choose the architecture. All the photographs with tanks were from a cloudy day and all the images without. After the neural network’s training was complete. They collected one hundred images with a concealed tank. with and without tanks. only to find that on this round of testing. the US military wanted to use a neural network to analyse images and determine whether they contained a concealed tank. A tool used to solve heuristic problems. the network came up with apparently random answers. The US military then tested the neural network with a further set of one hundred images. A classic example is that in the 1980s. Fifty pictures of each set were set aside and the neural network trained with the remaining one hundred images with and without tanks. train the neural network.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Once a neural network’s training is complete it is hard to tell why it is coming up with a particular solution. then a further one hundred images without. requires largely heuristic construction and preparation. for obvious purposes. or even to verify the quality of the training.

4. the more accurate and in line with expectations and desires. Garbage out”. One of the main research areas is using neural networks to recognise diseases/conditions via scans. lies in their ability for generalisation.Current Applications of Neural Networks 2. Fortunately since neural networks are (as has been mentioned above) good at ignoring irrelevant data. covering as many relevant situations as possible. This is generalisation and it is far more useful than it at first appears and similar from a very high-level perspective to how the human brain and brains in general function. “training” and “testing”. The real strength of neural networks. this has to be done within reason however. the eventual result will be. The more time spent training the neural network on quality data sets. in the opinion of the author. Data sets for training neural networks generally separate into two sub-sets. This drawback is however one that is generally universal to computing (and many other areas) often referred to as “GIGO” or “Garbage in. the usual best approach is to simply feed them as much data as can be obtained. 2.4 . The network training first takes place using the appropriate “training” set. Page 17 of 225 . Neural networks make it more apparent however because (as in the previous US Military example) it can be difficult sometimes to decide what constitutes good data.Neural Networks in Medicine There is a large amount of research now covering the applications of neural networks in medicine.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Another possible criticism is that neural networks are extremely dependant on the quality and content of the original training data set.1 . Testing may then take place with the “testing” set to determine accuracy with unseen data. or it does exacerbate the problem with the time spent training the neural network.

You can also use the technology to model parts of the human biological system in order to better understand it.Neural Networks in Business Neural networks. thus supplying information more or less directly linked to the airlines main income. mortgage screening.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 This is particularly suitable work for neural networks because you do not need perfect examples of how to recognise the diseases and conditions. A back-propagation neural net is integrated with the airline marketing tactician which monitored and recommended on booking for each flight. However. Businesses have used neural networks in the past for applications such as assisting the marketing control of airline seat allocations in the AMT (Airline Marketing Tactician). can fit very well into most business situations. a clear problem with this approach is that when using a neural network it would be difficult to tell how a neural network reached its conclusion. 1997) 2. being good at analysing patterns and predicting future trends. merely a variety of scans covering all possible permutations of the disease or condition. 1997) Page 18 of 225 . (Stergiou & Dimitrios. The main reasons for using a neural network to model the human cardio-vascular system are that a neural network is capable of adjusting itself to be relevant to a particular individual and also to adapt by itself to changes in that individual. Credit scoring.4. for example the human cardio-vascular system. In the 1980s.2 . research on using neural networks to diagnose diseases took place. and assessing borrowers have all incorporated neural networks as an integral part of the system. (Stergiou & Dimitrios.

4 . (relatively speaking) involving the following steps (Robot Control. The position.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 2.4. compensating automatically for wear and tear on Page 19 of 225 . 1999) 2. When a robot arm is one hundred percent accurate that is a simple task. 2008): • • • • Forward Kinematics Inverse Kinematics Dynamics Trajectory generation The ultimate goal of course is to position the “effecter” (I. a robotic hand or any other similar manipulator) in the appropriate position to grasp or otherwise manipulate an object or part of an object. velocity and acceleration of the object being estimated in those implementations by several neural networks using several of the most recent measurements of the object coordinates. Calculate the joint angles required for the arm to be in the appropriate position. (Marshall & Srikanth. 2008): • • • Determine the target coordinates relative to the robot.3 . Adjust the arm to the appropriate joint angles and close the effecter.Object Trajectories Predicting the trajectory of an object in real time is another area where neural networks have been utilised.e. A neural network allows the robotic arms to be very flexible in their operation and perform self-adjustments as time goes by. When controlling manipulators a neural net must deal with four major problems (Robot Control.4.Robot Control Controlling manipulators (such as robotic arms in a car manufacturing plant) is usually how robotics and neural networks connect.

Back-propagation 2. This rule essentially describes adjusting the weight using the difference between the expected and actual outputs. In order to use a back-propagation training algorithm you must have a non-linear activation function for your artificial neuron.Back-propagation overview In essence.1. allowing us to apply our learning algorithm. 2. A commonly used function is the sigmoid function as described in section 2.3. Adjust the weight of the neurons to minimise the responsibility. 2. based on the strength of the weights connecting them to the output neurons. 2008). Assign responsibility to each of the neurons in the previous level. The advantage of using a function like this is that it allows us to differentiate how close we came to the correct result. The sigmoid function is in the source code at 9.2 -The Artificial Neuron. Adjust the weight of the output neurons to lessen the error. This is far preferable to having to perform lengthy and expensive re-calibrations on any regular basis (Robot Control.1 .2 .1 -“Network.5.Back-propagation in depth In a single layer neural network (or with a single node) the perceptron rule is applicable.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 themselves.1.5 .cs” lines 130-133. the disadvantage of a differentiable function is that it is by the use of these functions that local optima become apparent in the search space.5. Page 20 of 225 . training by back-propagation involves the steps: • • • • • Present a set of training data Compare the network’s output to the desired output at each output neuron and calculate the error.

2 -The Artificial Neuron) an alteration can be made to the perceptron learning rule allowing this difficulty to be overcome. Back-propagation for the Uninitiated.Altered perceptron learning rule (Matthews.Altered delta calculation rule for the output layer (Matthews. such as the sigmoid function (see 2. 2002): ∆ = ℎ = − Equation 5 . 2002) In this equation “w” and “x” are as before. = 1− Equation 4 . as the effect that the change in the weight will have on the rest of the network is a missing factor. Provided you are using a differentiable function. 2002) Therefore.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 ∆ = ℎ = − . Equation 3 – The delta rule (Matthews. The sigmoid function differentiates very neatly. Back-propagation for the Uninitiated. Page 21 of 225 . Back-propagation for the Uninitiated. 2002) Calculating the hidden layer deltas is a little more complicated. Assuming the sigmoid activation function is in use. this rule has no positive effect in a multi-layer neural network. ℎ : = [1 − ] . The other two values “di and yi” are the actual and desired outputs. respectively. Back-propagation for the Uninitiated.1. Back-propagation for the Uninitiated. the output layer nodes training function can be written as: = 1− − Equation 6 . The effect that the change will have on the following neurons in the network is an extra factor. “n” is the learning rate of the neural network.differentiated sigmoid function (Matthews. 2002) However. the following alteration to the perceptron learning rule would serve our purposes (Matthews.

The FANN neural network library (Nissen. 2. 2002) As demonstrated above. 2.Altered delta calculation rule for the hidden layer(s) (Matthews. it is possible to “breed” solutions to problems. Back-propagation for the Uninitiated.7 . The FANN library was decided upon for several reasons. it is fast. 1859) very neatly avoids one of the larger problems involved in software design: how to specify in advance all the possible permutations of the original problem and how the program should react to those permutations. open-source (allowing me to alter it if I needed too.Genetic Algorithms Genetic algorithms attempt to copy natural laws to a certain extent in order to apply a random search pattern to a defined search space.3 . Most organisms evolve by means of sexual reproduction and natural selection (Darwin. against which to compare the genetic algorithm trained neural network.Back-propagation library 2. By using similar techniques to natural selection. in order to apply this learning mechanism you start at the output of the network and work your way towards the inputs.6 . although as it turned out I did not). FANN) The FANN library (see (9. 1859) Page 22 of 225 . to calculate the delta for a hidden layer requires the deltas for the following layers.3 -Libraries Researched) for other libraries investigated) implements a back-propagation solution and this library will be used in the project to implement a feed-forward back-propagation neural network.5. in a language that can be linked to C# fairly easily and it also supports cascade neural networks. easy to implement. Therefore.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Equation 7 . saving me finding a separate library. Natural selection (Darwin.

simulating processes of biological evolution.1 . At first. Also in 1975. By the early 1980s. By 1962. the publication of the book “Adaptation in Natural and Artificial Systems” occurred. During the process chromosomes line up together and then cross over partway along their length. across a range of subjects.7. selection and crossover. swapping genetic material. mostly designed to help evolutionary biologists’ model aspects of natural evolution. In 1965 one Ingo Rechenberg. although this technique involved no crossover. genetic Page 23 of 225 . Mutation plays a part in this as well. however.History of Genetic Algorithms Genetic algorithms first appeared on computers in the late 1950’s and early 1960’s. built on papers both by himself and by other researchers. during this process. 2. This book presents the concepts of adaptive digital systems using mutation. genetic algorithms were becoming more widely used. this mixing leads to much faster evolution than if off-spring simply copied the genetic material as it is in the parents. this application was largely theoretical. then of the technical University of Berlin introduced a technique he named evolution strategy. occasionally causing the genetic information. but this variety helps to prevent stagnation in the gene pool. This could result in either an effective or defective solution. to alter slightly in a random way. including noisy and discontinuous data. many researchers had independently developed evolution-inspired algorithms for function optimisation and machine learning but did not get much follow up to their work.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Sexual reproduction ensures mixing and recombination of genes. a dissertation by Kenneth De Jong established the potential of genetic algorithms by showing that they perform well on a wide variety of data. In 1975. or indeed multiple genomes in a population. by the author John Holland.

William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 algorithms soon moved into commercial territory and nowadays help solve many problems in many different areas.2 . or further up the hierarchy.7. in the same way that the average response of a relatively small percentage of the population of any given country. abandon that search. more conventional.2. 2004). (Marczyk.7. This leads to the exploration of a large proportion of the search space. they are very well suited as a means of exploring search spaces too large to be effectively searched in a reasonable amount of time by other. with processing power exponentially dedicated to better areas as the exploration progresses.Advantages The primary advantage of genetic algorithms is their inherent parallelism. proceeding with other more likely candidates. The real advantage of parallelism however is that by evaluating the relatively small search-space that it does. methods.Advantages and Disadvantages of Genetic Algorithms 2. These are “non-linear” Page 24 of 225 . Thanks to the parallelism that is a genetic algorithms main advantage. 2004) 2. This is “Schema Theorem” and allows a genetic algorithm to move towards the searchspace with the most promising individuals in a timely fashion and then select the best member of that group (Marczyk. Many other algorithms are largely serial and can only explore a search-tree so far in any one direction before abandoning all progress and starting again from the beginning. can be used to fairly accurately predict national trends. a genetic algorithm is implicitly evaluating a much larger group of individuals.1 . Genetic algorithms effectively explore many different branches of the tree at once and when a certain branch turns out to be non-optimal.

William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 problems. In infinite or very large search spaces. This non-linearity results in a huge increase in the search space. Genetic algorithms have no prior knowledge about the problem they are trying to solve. Many real life problems are not like this. 2004). due again to the parallelism of their approach. or never contained in their search space in the first place. crossover allows promising solutions to share their success and mutation allows random changes in the local search space of a given solution. due to their strength in navigating large search spaces (Marczyk. such as crossover and mutation. Selection allows the pruning of the least promising searches. Without crossover. or merely very good local optima. A third advantage of genetic algorithms is that they do not tend to be easily trapped by local optima. A good example of this is the concept Page 25 of 225 . each component’s fitness is individual and any improvement to individual component's fitness is necessarily an improvement to the whole. selection and mutation a genetic algorithm is metaphorically similar to a large number of parallel depth first algorithms. they tend towards non-linearity where altering one component's fitness positively may make other components less fit and cause a ripple effect across the whole system. they produce random changes to their candidates and then use the objective function to determine whether those changes are positive or negative overall. the random nature of the various starting points of the initial population and the other methods they employ. genetic algorithms tend to give good results compared to other search strategies. However. it is hard to know whether we have reached the global optimum. This allows them to discover solutions that other algorithms may have over-looked. each searching their own space for the best solution. In a linear problem. making genetic algorithms an effective way to search them. One of the largest strengths of genetic algorithms can at first glance appear to be their largest weakness.

rediscovered by genetic algorithms. Page 26 of 225 . a comparatively smaller segment of the search space will be covered and the eventual solution may take longer to reach or never be reached as a result. the investigation may cover too little of the search space to find the optimum solution (Marczyk. however that is only one point. the global optima is at the far right. therefore no gradual increase can lead you to it. where the actual code of the program forms the genomes in the genetic algorithm and it can swap portions in or out. then genetic algorithms (and most other search techniques) are no better than a random search for finding the solution.7. 2004).2 . integer.Disadvantages One of the disadvantages of a genetic algorithm is the necessity of representing the problem the genetic algorithm is trying to solve. population size. but denied a patent for several years because it ran counter to established beliefs (Marczyk. Generally. with each number representing some distinct part of the solution (Marczyk.2. in a form that the genetic algorithm can use. or real-valued numbers. if the solutions to various problems are in areas of the search space that the fitness function would not consider likely. etc. the system will never converge towards a suitable solution. the genetic algorithm may end up solving an entirely different problem from the originally intended one. A second disadvantage of genetic algorithms is that the fitness function is crucial to the development of a suitable solution for the problem. A third issue is setting the correct mutation rate. genetic algorithms use strings of binary.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 of negative feedback. Likewise. 2. For example in the diagram below. If you have a poorly written fitness function. A further problem is that fitness functions can be deceptive. if the size of the population is too low. Another method is to use genetic programming. 2004). or adjust them. there is no slope leading to it. however if it is too low. 2004). If the mutation is too high.

There are various methods for solving this. including sigma scaling and Boltzmann selection (Mitchell. 1996). but altogether easier to find.local optimum (units are undefined in this case). 2004). Page 27 of 225 . lowering the diversity of the population and resulting in genetic algorithms possibly falling into the local optima that the mutation represents. This is because analytical solutions usually produce more accurate results faster than a genetic algorithm (or any heuristic method) is capable of (Marczyk.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 you either have the solution or you don’t. A further problem is premature convergence: this occurs when a mutation early in the process produces a large leap in the fitness of that particular genome. unless by random chance it discovered the point with the most optimum solution. 900 800 700 600 500 400 300 200 100 0 0 2 4 6 8 10 12 Figure 4 . In this example a properly constructed genetic algorithm would be likely to settle on the marginally less optimal. which will then reproduce abundantly. where analytical solutions exist they should be take precedence over genetic algorithms. local optima in the centre of the diagram. Finally.

1 . There are methods to try to avoid this problem when using back propagation. however. Genetic algorithms also have the benefit that they do not place any restrictions whatsoever on the architecture of the neural network. 1995).Determining Weight Values with Genetic Algorithms Genetic algorithms (as described in 2.7 -Genetic Algorithms).7 -Genetic Algorithms) are good at avoiding being stuck on local optima due to their searching several regions of search space simultaneously. Page 28 of 225 . a neural network representation within a genetic algorithm is a concatenation of all the weighs in the neural network. 1992). 2. the only information genetic algorithms require to train a neural network is an objective function (as described in 2.2 .William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 2.8 . these are not a complete solution (Blum. to test the appropriateness of a given genome.Representation of a Neural Network within a Genetic Algrithm As a rule. since you are merely increasing the size of the genomes being worked with. however the method of choice for training neural networks is usually back propagation (see 2. In addition. 2. The error surfaces associated with these problems tend to be highly variable and contain many local optima. such as adding a “momentum” to the algorithm.8.Training Neural Networks with Genetic Algorithms Training neural networks is a form of problem solving. involving finding the optimum values for a set of real numbers (connection weights) which will produce the least error in the results from the neural network.5 -Back-propagation) which is a gradient descent algorithm and as such can be easily trapped in local optima (Branke. to try to help it move beyond local optima. not altering them besides that in any way.8.

William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Crossover can disrupt genes far apart. These are all connected. however representing a large number of real valued numbers as a binary string. 2. Because of this it is helpful to keep functional units close together (I. 1995). it is also possible to use a genetic algorithm to evolve a structure for a neural network. The average neural network has an input layer.e.3 . but is less likely to do so for genes close together. side by side). Page 29 of 225 .Using Genetic Algorithms to Determine Neural Network Structure As well as using a genetic algorithm to evolve the weight set for a fixed-structure neural network. one or two hidden layers of an indeterminate number of nodes and one output layer. a nodes input weights and the nodes of a layer. to form a feed forward network. outputs to inputs. place a neurons input and output weights. One of the most important decisions when deciding how to represent a neural network within a genome is whether to use binary strings or real number strings. can lead to very large strings (thousands of bits) (Branke.8. The problem at hand dictates the input and output layer structures. Standard genetic algorithms use binary strings.

If neural network architecture is too simplistic for the problem at hand. this is not necessarily the most efficient form for a network for any given problem and the trial and error method is an inefficient method for determining the correct architecture for an optimally performing neural network for the problem at hand. given a specific problem.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Figure 5 . There is also no method to check how optimal your solution is (Branke.A feed-forward neural network This architecture is then altered (number of nodes in each hidden layer. there is currently no method by which to establish. However. a neural network topology to deal with that problem. On the other hand. the neural network will never learn to solve the problem (or come close enough to suggest a solution). a neural networks generalisation is its main strength and therefore you do not wish to lose it. Page 30 of 225 . 1995). number of hidden layers) by trial and error and intuition to establish a neural network that performs well for the problem at hand. the neural network will learn fast but will generate too specific a solution and will not generalize between similar inputs. As mentioned in section 1.0. if the neural network architecture is too complex. Although network architecture plays such a large role in network performance.

3. thus meaning that in order to remove a node completely.8. a representation is required which can accommodate all networks that will work for the problem but none that will not. Low-level encodings specify each connection individually. direct.Representing neural network structure Representing the structure of a neural network within a genetic algorithm is not as straightforward as representing the weight values of a neural network within a genetic algorithm.1 . 1995). or low-level encoding and the weak. There are two methods for representing neural network topologies within a genetic algorithm. whereas high-level encodings group together connections and/or nodes. you must remove all connections to or from that node completely.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 2. In direct encoding. thus eliminating the searching of meaningless space and by definition. In one version of high level encoding of a neural network. In cases where a genetic algorithm determines the number of hidden layers and the number of nodes per hidden layer and the network itself is interconnected fully. Back propagation learning rate is also a parameter for one of these areas. Ideally. 1995). being able to contain the most optimum solution for the problem at hand. or high-level encoding (Branke. the network is divided into areas and for each area the number of nodes and the density of connections to other nodes is determined (Branke. indirect. an exception is possible. the strong. the genetic algorithm represents the neural network by means of its connections. Page 31 of 225 .

will not ever succeed in learning the problem.2 neural networks that are too large will solve a problem fast but will not generalize well and therefore their usefulness is limited severely. This can lead to human bias entering the system and affecting the outcome. and the moving target problem (Fahlman & Lebiere. it is impossible to know whether these large and complex networks may be the most suitable way to solve the problem at hand. 2. This could be useful. as the size of the neural network increases. 1991). These mapping methods are therefore more suitable for small networks or networks with a small number of connections.Cascade Correlation The cascade learning architecture is a method by which the training algorithm builds the neural network as it proceeds. 1995). Page 32 of 225 . a balance between speed of learning and generalisation exhibited in the trained network is necessary.9 . This section largely drawn from information in: “Curved trajectory prediction using a self-organizing neural network” (Branke. As mentioned in section 5. The Cascade-Correlation Learning Architecture. Neural networks that are too small on the other hand. by limiting the potential size of the networks evolved and preventing the generation of overly large and complex networks but on the other hand. Therefore. which is obviously undesirable. These are the Step-size problem. A problem with high level encoding is that regular networks are favoured but not every network is equally probable. The algorithm attempts to solve the issues associated with back-propagation that produce slow learning.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 One of the reasons why alternatives to low-level encoding arose is the exponential rise in the number of connections.

1988). There is also a bias input. 1988).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 The system begins with a basic neural network of inputs and outputs (numbers of each dictated by the task as usual) and no hidden layer. An Empirical Study of Learning Speed in Back-Propagation Networks. After a number of training cycles pass with no substantial improvement (the precise number being determined by a user-supplied “patience” parameter). = −` −` . Page 33 of 225 . If a trigger occurs for the addition of a unit. with the aim of getting as close to a solution as possible. The “candidate” nodes take the neural net inputs. and “Eo”.2 -Back-propagation in depth) or the quick prop rule (Fahlman. The Cascade-Correlation Learning Architecture. 1991). or if the networks output is of sufficient quality the training is ended. the quantities ‘`v ’ and ‘`Eo’ are the values of v and Eo averaged over all patterns (Fahlman & Lebiere. the candidate unit’s value.2 -Back-propagation in depth).The definition of 's' given that 'o' is the network output at which the error is measured. The Cascade-Correlation Learning Architecture. 1991). 'p' is the training pattern. Equation 8 . In order to maximise "s".5. the partial derivative of "s" with respect to each of the candidate unit’s incoming weights "wi" is a necessity. or sometimes the quick prop rule (Fahlman. plus the outputs of any previously created units. 'v' is the candidate unit's value. It then trains the adjustable weights using the perceptron rule (as described in 2. and attempt to maximise “s” the sum overall output units “o” of the magnitude of the correlation between “v”. the initiation of a number of “candidate” units takes place and trained using either the perceptron rule (2. The output units may use either a linear or differentiable activation function. An Empirical Study of Learning Speed in BackPropagation Networks. the residual output error at “o” (Fahlman & Lebiere.5. with a value of one. either an extra node is added. The connection between each input and output is via an adjustable weight.

= . where ‘oo’ is the sign of the correlation between the candidates value output ‘o’. The connection of the best of the candidates to the neural network as it stands takes place. Once the correlation on a candidate has reached an acceptable level. Adjustable weights link the outputs for this node and all the output nodes inputs.The partial derivative of ’s’ (Fahlman & Lebiere. After these computations are complete. Equation 10 . ‘f'p’ is the derivative for pattern 'p' of the candidate unit's activation function with respect to the sum of it’s inputs and ‘Ii. The training iterations for these adjustable weights in the main network then continue.The partial derivative of ‘s’ with respect to each of the candidates incoming weights ‘wi’. −` ′ . a gradient descent to maximize "s" can take place. and the inputs for that node are frozen. The Cascade-Correlation Learning Architecture. 1991).p’ is the input the candidate receives from unit ‘i’ for pattern ‘p’. or the completed training epochs reaches a maximum level.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Equation 9 . Page 34 of 225 . Expansion and differentiation of this calculation can then take place. .

The Cascade Correlation Learning Architecture. 1991) Cascade-Correlation 1991). A first node has been added. Page 35 of 225 . 1991) Cascade-Correlation 1991). The circular connections are adjustable weights (Fahlman & Lebiere. The square econd connections are locked weights (Fahlman & Lebiere.The second state of a cascade correlation neural network.The initial state of a cascade correlation neural network. Figure 7 . The Cascade Correlation Learning Architecture.William Sayers 05025397 Genetic Algorithms and Neural Networks Neural Milestone 3 Figure 6 .

The third state of a cascade correlation neural network.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Figure 8 . Page 36 of 225 . with two nodes added (Fahlman & Lebiere. 1991). The Cascade-Correlation Learning Architecture.

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3

2.10 - C# User Interface Programming
C# uses the Winforms API to create user interfaces. It allows you to “draw” objects on a form in a visual manner to design your layout and then access and alter properties and subproperties of those objects in your programming code. A separate thread for each Window in Winforms exists, in order to facilitate performance on modern multi-core machines, although it does make it more difficult to move data between the forms. A method using “delegate functions” (similar to “function pointers” in C++) and “events” (Lysle, 2007) is the method I will use. Another challenge is making the neural network training independent of the user interface, in order to avoid the interface freezing whilst training a neural network. The solution I plan for this is to use the background worker object provided with C#, which allows you to assign a delegate function to be your background task, and will trigger events upon completion, for which event handlers will be executed (Microsoft, BackgroundWorker Class, 2009). Making a C# user interface interact with C++/CLI dll’s simply involves adding a reference to the dll in your project, via the Visual Studio® user interface.

2.11 - Writing C++/CLI Wrappers for Unmanaged C++ code
The cleanest method of connecting a native C++ or C dll or piece of code to a managed C# interface is via a C++/CLI bridge class. The VC++ compiled can swap between managed and unmanaged code on the fly whilst compiling and allows you to have both managed and unmanaged code (in C++) as part of the same project. Thus, a managed class can communicate with the unmanaged class (performing appropriate marshalling of the non-primitive data types (Microsoft, Using C++ Interop (Implicit PInvoke), 2009)) and the C# code can then access it directly.

Page 37 of 225

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3 Although it is possible for a C# program to interact with an un-managed dll directly, this method is less error-prone and more likely to be robust and stable, as well as making for cleaner code.

2.12 - Application of Research
The research presented here will help to produce the deliverables for milestone two and milestone three of my final year project: I plan to combine C# for the user interface and C++ for the neural network back-ends in order to create an efficient, powerful and fast solution. Section 2.11 -Writing C++/CLI Wrappers for Unmanaged C++ code, covers the combination of C# and C++ so I will not bother to cover it here, I plan to use the DLL method, with one DLL combining managed and unmanaged C++ linked into my C #user interface code. The user interface will then have methods to adjust parameters such as the genetic algorithms population size, number of generations, etc. In order to test and see what kind of genetic algorithm learns and develops fastest. It will also allow you to specify the number of back-propagation iterations. The use of this application will be in evaluating the difference in performance between a genetic algorithm, back-propagation algorithm and cascade correlation algorithm as neural network training algorithms.

3 - Design
The purpose of this application will be to allow the user to run a back-propagation trained neural network, or a genetic algorithm trained neural network and present the results of the network in such a fashion as to allow analysis and comparison of the training methods and their suitability in different situations.

Page 38 of 225

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3

3.1 - Program Requirements
The program must be able to: • • • • • • Run a genetic algorithm on a feed forward fully interconnected neural network to train the network. Run a back propagation algorithm on a similar network to train the network. Run a cascade algorithm on a similar network to train the network. Allow the user to set appropriate variables (see 3.4.2 -New Network Wizard) to adjust the execution of the learning algorithm selected for the neural network. Allow the user to select which learning algorithm they wish to use. Display the results of running that algorithm on a neural network in a meaningful fashion.

3.2 - Design of the Class Structure
Before I discovered a simple to program and effective design, I tried and discarded two designs. I eventually used the third design and it is that third design documented here. The two prior designs to this involved representing neural networks via objects and compositions of objects in vectors, which lead to flexible, but slow and overly complex code upon experimentation. The final solution, the solution documented here, is a far simpler one (in terms of the neural network representation, it is far more detailed in the other areas than the initial designs were). This solution loosely follows a model, view, controller architecture. The solution enables easy replacement of sections of code as long as the interfaces and outputs remain the same. For more information on the class structure, see section 9.2 -UML Class Diagram.

Page 39 of 225

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3

3.3 - Linking C# code to managed dll’s
Linking C# code and managed dll’s is a simple process (in Visual Studio) as explained in (2.11 -Writing C++/CLI Wrappers for Unmanaged C++ code) of selecting the references section in the C# project in question and adding the dll you wish to refer too. Then you can simply add a “using” directive to use the dll namespace and create, destroy, and use the classes within as if they were native C# classes. This can be seen in (9.1.2.6 “Algorithms\BackProp.cs”) and (9.1.2.4 -“Algorithms\CascadeCorrelation.cs”) with the “using FANN_Wrapper;” directive. This is superior to the direct C# to unmanaged code linking that used in milestone two, since it makes for cleaner and thus more easily maintainable code and the code is more robust in this fashion. The robustness comes from the fact that C++/CLI is far better at interacting with unmanaged C++ code than C# is. Indeed – they can even be in the same source file and compiled into one object. Therefore, by using a C++/CLI wrapper to interface to the unmanaged C++ and then interacting with the C++/CLI wrapper/s errors are less likely to arise.

3.4 - Design of the User Interface
The construction of the user interface utilised the Winforms based tools in visual studio 2008 (see 2.10 -C# User Interface Programming). I also used the ZedGraph control (ZedGraph) to display the graph of my results.

3.4.1 - Main Form
The main form is the crux of the application (as is default in C# Winforms programming) and if the main form is closed, all other forms and classes are disposed and the application terminates.

Page 40 of 225

Figure 9 – Network and output tab design (Main form) 3. One display will show network details and one display will show the training and testing output.Mean Squared Error Graph tab The mean squared error graph tab is a simple tab.4.4. Three buttons will therefore be present. Page 41 of 225 .1.1 .1.2 -New Network Wizard) a method to train the current network and a method to test that training.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 3. which simply needs to display the graph of the mean squared error over the current epoch/generation.Network and Output tab This tab needs a method by which to initialise the adding of a new network (3.2 .4. along with two displays.

to show the training data and to show the testing data. Figure 11 .3 .William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Figure 10 .Mean squared error tab design (Main form) 3.Dataset tab This tab needs two display sections.Dataset tab design (Main form) Page 42 of 225 .1.4.

William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 3. Verification of all data takes place before the form passes control back to the main form and triggers a main form event with a parameter that takes a specially constructed class to transfer the data from the wizard. The user can progress through the steps via the “previous” and “next” buttons or via the tabs at the top in whatever order you choose.4. Page 43 of 225 .2.4.2.Design of the training algorithm tab (Wizard form) 3.Learning Algorithm tab On the learning algorithm tab you select the learning algorithm for the neural network you are creating.7 Genetic Algorithms.1 -Learning Algorithm tab (if no algorithm has yet been selected.New Network Wizard The new network wizard is the form that appears when the user selects the creation of a new network.9 -Cascade Correlation).2 .4.5 -Back-propagation.2 . 3. It allows you to move in a logical progression through the set up process of a new neural network and associated training algorithm.1 .2.4. 2. Figure 12 .Network Settings tab This tab allows you to set appropriate settings for the network and training algorithm you have selected in 3. out of three choices previously mentioned (2. 2. the default is genetic algorithm).

both to protect the main form from data inputs that may cause training problems and to provide a visual indicator of work taking place. Figure 14 .3 .4. Page 44 of 225 . or to choose custom data sets from your hard disk for training and testing data (appropriately formatted in the FANN style).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Figure 13 .Dataset selection tab design (Wizard form) 3.Design of the network settings tab (Wizard form) 3.Dataset tab The dataset tab allows you either to select from two pre-setup datasets.Working Form The working form is merely a small display shown as a dialog (lock out access to displaying form) which shows whilst a neural network is training.3 .2.4.

one tab with XOR information and one tab with Fishers Iris Data information.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Figure 15 . Clicking the datasets button or the licenses button displays the appropriate information.4. project supervisor.4 .1 .Datasets display Clicking the data sets button displays a tabbed interface.About form design (About form) 3.About Form The about form displays information about the application (author.4. Figure 16 . clicking the same button again hides that information.Working form design (Working form) 3.4. Page 45 of 225 . displays licensing information and displays information on the pre selected datasets. project description).

Figure 18 .Dataset display design (About form) 3.Licenses display design (About form) 4 . depending on your selection.2 .Implementation The implementation is largely in C#. Page 46 of 225 .4. FANN) library written in C and linked as a static library to the managed FANN_Wrapper dynamic link library.4.Licenses display Clicking the licenses button displays a small information panel with selection methods. with small portions in C++/CLI and the back-end FANN (Nissen. it will display the appropriate license.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Figure 17 .

William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 4.Main Form The main forms code is in section 9.1 .1.1 . mostly via the visual design tools found in visual studio 2008.1 . 4.User Interface Implementation The implementation of the user interface is mainly in C#.1.1. 4.cs”.Network and Output tab Page 47 of 225 .1.Network and Output tab Figure 19 .1.2 -“FrmMain.

1.Mean Squared Error Graph tab Figure 20 .William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 4.1.Mean squared error graph tab Page 48 of 225 .2 .

The dataset tab Page 49 of 225 .1.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 4.Dataset tab Figure 21 .3 .1.

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3

4.1.2 - New Neural Network Wizard
4.1.2.1 - Learning Algorithm tab

Figure 22 - Learning Algorithm tab

Page 50 of 225

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3

4.1.2.2 - Network Settings tab

Figure 23 - Network Settings tab

4.1.2.3 - Dataset tab

Figure 24 - Dataset tab

Page 51 of 225

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3

4.1.3 - About Form

Figure 25 - About form, main view

Page 52 of 225

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3

4.1.3.1 - Datasets View

Figure 26 - About form, Datasets view

Page 53 of 225

3. Licenses view 4.4 .5 .1.Passing Information between Forms Passing the data between the wizard form and the main form takes place as previously described (2.About form.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 4.2 .1.Working Form Figure 28 .Licenses View Figure 27 .Working form 4.10 -C# User Interface Programming) via events and delegate functions.1. Page 54 of 225 .

The “NetworkUpdatedEventArgs” class is a storage class for passing data (see 9. Events. The declaration and association of a delegate and an event takes place in the form that needs to generate an event (9.1.1.2 -“FrmMain. The trigger of a new event takes place in the Wizard form when you are on the last tab and the user clicks “next” (see 9.2 “FrmMain.1. the processing of the class sent from the Wizard form takes place (see 9.1. 2009).1 -“NetworkUpdatedEventHandler”).1.cs”. BackgroundWorker Class.3 -“Wiz1.1. This leads to a user interface that is much more accessible (Microsoft.cs” lines 145-163).1.1. In the event handler in the main form. lines 29 – 31).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 An event is a method of notifying a listening object that something has occurred.Keeping the User Interface Active whilst Processing is Occurring The user interface delegates the processing to a background worker class. The two-event handling functions are at 9.1.1.1.1. Then the runworkerasync function executes and then shows the working form.6 . (See 9.cs”.2 -“FrmMain.2 -“FrmMain. The declaration of a new event handler in the main form takes place prior to displaying the new neural network wizard form (see 9. Page 55 of 225 .1.1.2 -“FrmMain. lines 306-353.3 -“Wiz1. lines 266-281).cs” lines 120-129).cs” lines 70-119). the background worker class is instantiated and supplied with the event handler references functions (see 9. are associated with a delegate function to respond to that event (Lysle. 4. Firstly. when instantiated.cs”.1.1. 2007).cs”.1. lines 29-44).

The neural network specific functionality added to the network model in the network specialisation class GANetwork (see 9. and didn’t suffer from the memory leaks (being managed code) so that’s the version that exists in the application today (see 9.1. I decided to try re-implementing it in C#. • • The sorting implementation in GANetwork.NET sort methods in the sorting part of the genetic algorithm). There is an unmanaged C++ wrapper (9.cs” for the C# implementation) The main points to note in the code as it stands are: • The crossover (single-point) and mutation (random increment or decrement) implementations (see 9.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 4.NET built in functionality (see 9. Once this implementation was complete.5 -“Algorithms\GeneticAlgorithm. it was both faster than the C++ implementation (due mostly to my use of build in .1 .4.3 Page 56 of 225 .cs” lines 51-126) 4.2.2. after having used C# extensively in the UI I felt more confident about the performance loss and the use of the programming language itself. when the C++ implementation of genetic algorithms that I wrote functioned.2 . but had issues with memory leaks and was slow.cs” lines 176204).2 -“GANetwork.Genetic Algorithm Implementation Although the original design was to implement genetic algorithms in C++. and the Dll linked to this to allow access to the functionality.5 “Algorithms\GeneticAlgorithm. FANN) which is a C programming language based neural network library. The compilation of the FANN library is static.cs that links into the .cs” lines 32-44).1.3.2 -“GANetwork.1. Therefore.2.3.1.Back-propagation and Cascade Correlation Implementation The FANN library implements both cascade correlation and back propagation (Nissen.1.

In the initialisation routines (lines 28-124).3. will also be the coordinates for its corresponding weight (lines 17-19). layers.cpp”. the initialisation of all the arrays to their correct sizes takes place and then the initialisation of the weights to a random double between zero and one.2 -“FANN_Wrapper. 9. and nodes) the eventual implementation turned out to be basic yet effective (9. 9. and the outputs.1 -“FANN_Wrapper. number of outputs) separates the data into two-dimensional arrays suitable to be passed to other functions.4.1.h”. 9. In the constructor (lines 41 – 77).2 -“DatasetParser. the weights. Page 57 of 225 . into arrays of strings.1. via a number of designated separator characters. which using the data from the heading (number of epochs of data. 4. number of inputs.cpp”) for the unmanaged C++ classes. which allows it to manipulate the FANN functions in the static library through this class.Neural Network Implementation Although two previous attempts made at programming a neural network. C# accesses the C++/CLI class directly.1. 4.4 “UnManaged_FANN_BackProp. the separation of the textual inputs takes place.cpp”) for the Cbased FANN functions in the DLL and a C++/CLI wrapper (9.Data Parser Implementation The data parser (9.3 .6 -“UnManaged_FANN_CasC. From there.4.cs”).1. via references to these objects passed down via the hierarchy (references are passed to avoid the necessity of passing entire strings down the hierarchy).1 -“Network.4 .1.1.5 -“UnManaged_FANN_CasC.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 “UnManaged_FANN_BackProp. these arrays form parameters for the other functions.h”.4. 9.cs”) parses data directly from the rich text boxes that they’re displayed upon on the main form. the inputs.4. which store respectively.4. The inputs and weights are structures so that the coordinates for an input. followed an object-based structure (network.1. Three jagged arrays are utilised.h”.2.

to run the output layer. 1. input layer.02 for Fishers iris data and the virus classification data. the dataset is too small to split.XOR In the XOR dataset. 5. Then entered the weights from the neural network program (by means of using the visual studio debugger to check variable values) into the spreadsheet and checked that the outputs coincided.4 -Datasets) (ten times for XOR.Testing Data Testing took place several times for each algorithm and dataset (see 9.1 . 5 . so repetition is instead used. and creating a neural network with a similar structure on an excel spreadsheet. five times for Virus classification and Fishers iris data).01 for XOR and 0. I tested the neural network implementation by creating a neural network with structure 2. and hidden layer which are all called by the public “run()” function (lines 241-270). Page 58 of 225 . The target error was 0. As the program presented the data in each case. there are three functions. All nodes determine their outputs using the sigmoid function (lines 130-133). recording took place for analysis. The genetic algorithm and back-propagation neural networks each had four hidden nodes for each dataset.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 The main functionality of the neural network class can be found in the “internal” region (lines 126-235). 3.

00999799 0.009888967 0.1.009888967 0.00987211 0.00998959 0.009248076 0.009962559 0.009248076 0.00995716 0.00991907 0. Iterations (BP) 345 287 319 327 301 323 421 315 362 325 332.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 5.00983181 0.0099765 0.1 .1.00957518 0.5 Result MSqE (BP) 0.009829335 0.00998032 0.009375132 0. because of floating point inaccuracies converting between double precision floating-point numbers in my code and single precision in the FANN code.Back-propagation In the back-propagation testing data.0098402 0.00992291 0.006150786 0.009566951 0.00992081 0.009885499 Table 3 .00988649 0.5 Result MSqE (GA) 0.2 .0099113 0.Back-propagation testing data (XOR) Page 59 of 225 .00881867 0.00957518 0.009162436 0.009375132 0.00989974 0.00997699 0.009829335 0.009157809 Table 2 .009157809 Testing MSqE (GA) 0.009566951 0.009962559 0.00881867 0.009969323 Testing MSqE (BP) 0.00988863 0.009162436 0.00999069 0.Genetic Algorithms Iterations (GA) 591 407 500 466 499 534 675 557 357 939 552. the mean squared error from the network result and the testing mean squared error are different despite the data being the same.006150786 0.00990403 0.00990161 0.Genetic Algorithm testing data (XOR) 5.00998027 0.

but in the case of cascade training. At each node addition.86E-14 1. which is when the program records the mean squared error.02E-13 2.43545E-13 Table 4 .22E-13 1. the network is trained more after that point (output weight training).00798691 0.00863005 0. the call back function triggers.83E-13 2.Cascade Correlation In the cascade correlation data.02E-13 1. Nodes (CC) 2 2 2 2 2 2 2 2 2 2 2 Result MSqE (CC) 0.00889787 0. the testing result differs quite widely from the network result.00585159 0.18E-13 7.008039565 Testing MSqE (CC) 7.78E-14 1.1.00656719 0.99E-13 1.14E-14 2.0086449 0.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 5.00855044 0.32E-13 2.Cascade Correlation testing data (XOR) Page 60 of 225 .00992534 0.00573252 0.3 .00960884 0.

Graphs 1000 900 800 700 600 500 400 300 200 100 0 0 5 10 15 Iterations (GA) Iterations (BP) Nodes (CC) Figure 29 .002 0 0 5 10 15 Result MSqE (GA) Result MSqE (BP) Result MSqE (CC) Figure 30 .01 0.Number of iterations to achieve target (XOR) 0.1.012 0.4 .008 0.004 0.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 5.006 0.Mean squared error upon last training test (XOR) Page 61 of 225 .

008 0.01 0.012 0.The mean squared error at testing (XOR) Page 62 of 225 .006 0.004 0.002 0 0 5 10 15 Testing MSqE (GA) Testing MSqE (BP) Testing MSqE (CC) Figure 31 .William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 0.

0199972 0.0141406 0.0143806 0.Genetic Algorithms Iterations (GA) 1928 1834 1277 2835 8716 3318 Mean squared error (GA) 0.019775896 0.01956255614 Testing mean squared error (GA) 0.019719802 0.019751257 0.0193595 0.014144 0.018789438 0.0199892 0.0189456 0.01950238 Testing mean squared error (CC) 0.3 .William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 5.1 .0199918 0.0199916 0.0140669 0.019763877 0.019751257 0.0195853 0.Cascade Correlation Nodes (CC) 2 2 2 2 2 2 Mean squared error (CC) 0.Genetic algorithm testing data (Iris data) 5.Cascade correlation testing data (Iris data) Page 63 of 225 .0199719 0.0160925 Table 7 .019719802 0.Back-propagation Iterations (BP) 434 529 474 391 527 471 Mean squared error (BP) 0.01998996 Testing mean squared error (BP) 0.2 .0198416 0.Back-propagation testing data (Iris data) 5.014116 0.0141935 Table 6 .0164964 0.0196297 0.Fishers Iris data 5.019763877 0.0139982 0.2.0158679 0.2.0199999 0.2.019775896 0.2 .0143778 0.018789438 0.01956255614 Table 5 .

2.4 .019 0.0194 0.02 0.Graphs 10000 9000 8000 7000 6000 5000 4000 3000 2000 1000 0 0 2 4 6 Iterations (GA) Iterations (BP) Nodes (CC) Figure 32 .Mean squared error upon last training test (Iris data) Page 64 of 225 .0196 0.0198 0.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 5.Number of iterations to achieve target (Iris data) 0.0202 0.0188 0.0192 0.0186 0 2 4 6 Mean squared error (GA) Mean squared error (BP) Mean squared error (CC) Figure 33.

Virus Classification This dataset covers 61 viruses affecting several crops including tobacco.02 Testing mean squared error (GA) Testing mean squared error (BP) Testing mean squared error (CC) 0. cucumber.005 0 0 2 4 6 Figure 34. and others.01 0.025 0.The mean squared error at testing (Iris data) 5.015 0. which are the number of amino acid residues per molecule of coat protein.3 . There are 18 measurements on each virus. The dataset is in order: • • • • Hordeviruses x3 Tobraviruses x6 Tobamoviruses x39 Furoviruses x13 Page 65 of 225 . tomato.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 0.

019092293 0.8 Mean Squared Error training (GA) 0.0190612094 Table 8 .3.017043 0.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 5.0199621 0.1 .0194383 0.0199916 0.0196959 0.4 Mean Squared Error training (CC) 0.Cascade Correlation testing data (Viruses) Page 66 of 225 .Genetic Algorithms Iterations (GA) 1818 1421 896 11676 1068 3375.Genetic Algorithms testing data (Viruses) 5.01870197 0.01870197 0.017723052 0.017723052 0.2 .2 Mean Squared Error training (BP) 0.0111416 0.019861371 0.0185152 0.0347797 0.0391179 0.00640291 0.0350318 0.0197935 0.019927361 0.019927361 0.011585152 Table 10 .3.0434423 0.3 .0105845 0.0393106 0.3.00980208 0.03833646 Table 9 .019743 0.Cascade Correlation Nodes (CC) 1 2 1 2 1 1.019092293 0.Back-propagation Iterations (BP) 3275 10922 2800 3744 6295 5407.0190612094 Mean Squared Error testing (GA) 0.0165979 0.00979968 0.016183342 Mean Squared Error testing (CC) 0.Back-propagation testing data (Viruses) 5.01983722 Mean Squared Error testing (BP) 0.0195173 0.019861371 0.

Number of iterations to achieve target (Viruses) 0.02 Mean Squared Error training (GA) Mean Squared Error training (BP) Mean Squared Error training (CC) 0.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 5.01 0.4 .Mean squared error upon last training test (Viruses) Page 67 of 225 .015 0.Graphs 14000 12000 10000 8000 6000 4000 2000 0 0 2 4 6 Iterations (GA) Iterations (BP) Nodes (CC) Figure 35 .005 0 0 2 4 6 Figure 36 .3.025 0.

Comparisons 6.045 0.005 0 0 2 4 6 Mean Squared Error testing (GA) Mean Squared Error testing (BP) Mean Squared Error testing (CC) Figure 37 .04 0.01 0.03 0. Page 68 of 225 . Genetic algorithm solutions are slightly more accurate. versus backpropagations 333 iterations on average.1.Mean squared error at testing (Viruses) 6 .025 0.02 0. Genetic algorithms take around five hundred iterations on average to produce a workable solution when aiming for a mean squared error of below 0.XOR As regards to the XOR problem.Back-propagation and Genetic Algorithms 6.035 0. or possible inefficiencies in my programming. genetic algorithms and back-propagation offer similar performance in solving the problem.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 0. The back-propagation implementation seems to be slightly faster in real-time but this is possibly attributable to managed code inefficiencies versus unmanaged code.1 .1 .01.05 0. but to a negligible degree in this scenario.015 0.

2 .William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 6. The genetic algorithm solutions also held their effectiveness through to the testing data with extreme consistency.2.3 .Fishers Iris Data Genetic algorithms clearly found this a difficult problem to solve as opposed to backpropagation.XOR This problem was solved more effectively by the Cascade algorithm than by genetic algorithms.Fishers Iris Data Again. FANN)) this is around 300 epochs to solve the problem.2 epochs for backpropagation. 6. The back-propagation generated networks were also more accurate overall (in fact on this particular problem analysed the testing data more effectively than the training data).Virus Classification The solution of this problem was more effective by genetic algorithms than by backpropagation. With an average of 3318 generations for genetic algorithms versus 471 epochs for back-propagation this became clear.Cascade Correlation and Genetic Algorithms 6. genetic algorithms taking 550 generations on average to solve the problem to a suitable accuracy. whilst the cascade algorithm used 2 nodes to solve the problems.2.1. whereas the back-propagation solutions lost effectiveness on the training data.2 . At a maximum of 150 epochs per node added (the default (Nissen.8 generations versus 5407.2 . on this data set cascade training came to a satisfactory solution faster and more accurately than genetic algorithms. 6. The cascade algorithm also achieves a higher degree of accuracy. 6.1. with an average of 3375.1 . Page 69 of 225 .

3 .3 . Page 70 of 225 .Testing conclusions Genetic algorithms are a valuable and effective means of training neural networks.Virus Classification With this data set cascade training again came to a satisfactory solution faster and more accurately than genetic algorithms did on the same data. With more time.2. genetic algorithms fall down as a solution. more in-depth testing using more varied datasets a clearer comparison and possibly drawbacks associated with cascade training may become evident.Evaluation Overall. the unavoidable conclusion is that cascade training is the most effective of the three training methods examined. They are roughly as effective as back-propagation training and although back-propagation is faster for the most part. Compared to cascade correlation however. 6. I feel the project has been a success.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 6. although there are further investigations I would like to continue with given more time (and may continue with in my own time). solutions developed using genetic algorithms tend to hold their effectiveness through to the data set effectively. with the information presented and discussed here. however. 7 .

Build an application that is capable of training a neural network with a genetic algorithm and with a backpropagation system. Identify the most effective ways of training neural networks using genetic algorithms. and in which cases these methods are suitable. Using above application. and that allows manual adjustment of relevant variables.How many of the original objectives were achieved? The original objectives were as follows (9.5 -Objectives): Research the fields of genetic algorithms and neural networks.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 7.” With more time more advanced tests could be undertaken using larger data sets. evaluate training methods for neural networks.1 . as well as previous research. Page 71 of 225 . “Identify the most effective ways of training neural networks using genetic algorithms. The only objective that I feel could be improved upon is the final objective. Research current information on training neural networks using genetic algorithms. leading to a more solid conclusion and more information on the various strengths and weaknesses of the various learning algorithms examined. and in which cases these methods are suitable. Build a user-interface for the aforementioned application allowing for easy alteration of appropriate variables.

These could be resolved fairly simply (I believe) by implementing the iDispose interface in my managed code to allow it to be destroyed upon demand instead of waiting for the garbage collector. Due to small conflicts between managed and unmanaged memory.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 8 .Possible Improvements More advanced testing would be the biggest improvement that more time would make possible. the program occasionally causes an exception (most specifically when creating cascade or back-propagation networks in quick succession). Page 72 of 225 . Testing both with and without scaling and utilising different scaling methods. Some forms of testing that would be possible are: • • • • • Testing with datasets with larger numbers of inputs Testing with large datasets that attempt to train the network to distinguish between small differences in data. Testing with datasets that would take large amounts of time to train. Improve memory management.

Possible Training Algorithm Improvements With more time.2 . Having the back-propagation algorithm in C# similar to the genetic algorithm would also allow a more “apples to apples” comparison.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 8. is worth the possibility of reduced performance. experimenting with multiple point crossover and other crossover forms. Page 73 of 225 . for the reasons stated above. 8. • • Attempt to improve the speed of the genetic algorithm. Making the forms shown as dialogs. • Implement the cascade training algorithm in C# instead of C++. I would experiment with the following training algorithm improvements: • Improve the genetic algorithms crossover and mutation algorithms.1 .Possible User Interface Improvements The main improvements to the user interface that I would make if I had more time are: • • Increased number of variable options presented to the user (with appropriate defaults set). Implement the back-propagation algorithm in C# instead of C++. Contrary to my opinion before undertaking extensive usage of C# I now believe the reduced development time. allowing more to be undertaken successfully. but still minimize instead of blocking other computer programs. (the new network wizard and the working form) lock out the rest of the program. as well as implementing roulette wheel selection.

and neither is the code for the prototype constructed in milestone two. Page 74 of 225 . which this code is partially based upon.1 . I have not included it in the source code listing. The earlier attempts at genetic algorithm and neural network implementations are not included. so that this listing contains only code I have personally written for the project. This section also contains only the final draft of the code. Although there was another section of the program compiled from source (the FANN library).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 9 .Appendices 9.Source code The source code in this section is organised first by which part of the application it represents and then by what source file it originally resided in. The code contains line numbering for pinpointing the precise lines in question within the body-text when crossreferencing.

Controller.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 9. Page 75 of 225 .Classes { public class NetworkUpdateEventArgs : System.EventArgs { #region data /********************\ |* DATA *| \********************/ //Training Algorithm Identifier private int trainingalg.Linq.Generic.1.Text. System. System.1 .1. System. private bool _CC_ITerr. private double _CC_LearningRate.Collections. private int _CC_MaxNeurons. //Cascade Correlation private double _CC_Terr. //Dataset identifier private int _TrDataSet.1 .“NetworkUpdatedEventHandler” 00000001 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 00000018 00000019 00000020 00000021 00000022 00000023 00000024 00000025 00000026 using using using using System.1. private int _CC_Reports. namespace _05025397.View 9.

private double _BP_LearningRate. #endregion #region constructor /********************\ |* CONSTRUCTOR *| \********************/ //Constructor to accept all necessary data public NetworkUpdateEventArgs (int Algorithm. int GAPopSize.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000027 00000028 00000029 00000030 00000031 00000032 00000033 00000034 00000035 00000036 00000037 00000038 00000039 00000040 00000041 00000042 00000043 00000044 00000045 00000046 00000047 00000048 00000049 00000050 00000051 00000052 00000053 00000054 00000055 00000056 //Genetic Algorithm private int _GA_PopSize. double CCLearningRate. Page 76 of 225 . //BackProp private int _BP_EpochLimit. private bool _GA_ITerr. int CCReports. private double _BP_Terr. private int _BP_HiddenL. private int _GA_Mutation. private int _GA_HiddenL. private int _GA_GenLimit. private double _GA_Terr. private bool _BP_ITerr. private int _GA_Crossover. private int _GA_Reports. double CCTerr. private int _BP_Reports. bool CCITerr. int CCMaxNeurons.

double GATerr. _GA_GenLimit = GAGenLimit. _GA_Crossover = GACrossOver. int BPReports. _CC_ITerr = CCITerr. _CC_MaxNeurons = CCMaxNeurons. int GAMutation. int BPHiddenL. //Genetic Algorithm _GA_PopSize = GAPopSize. _GA_Terr = GATerr. double BPLearningRate. int GAHiddenL. //Cascade Correlation _CC_Terr = CCTerr. //Backpropagation Algorithm _BP_EpochLimit = BPEpochLimit.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000057 00000058 00000059 00000060 00000061 00000062 00000063 00000064 00000065 00000066 00000067 00000068 00000069 00000070 00000071 00000072 00000073 00000074 00000075 00000076 00000077 00000078 00000079 00000080 00000081 00000082 00000083 00000084 00000085 00000086 int GAGenLimit. int TrData_Set) { //Algorithm trainingalg = Algorithm. int BPEpochLimit. _GA_Mutation = GAMutation. _CC_Reports = CCReports. bool GAITerr. int GAReports. Page 77 of 225 . int GACrossOver. _GA_ITerr = GAITerr. bool BPITerr. _CC_LearningRate = CCLearningRate. _GA_HiddenL = GAHiddenL. double BPTerr. _GA_Reports = GAReports.

_BP_ITerr = BPITerr. } #endregion #region getters/setters #region cascade_correlation //Cascade Correlation public double CC_Terr { get { return _CC_Terr. } } public int GA_GenLimit Page 78 of 225 . } } public bool CC_ITerr { get { return _CC_ITerr. _BP_Terr = BPTerr. _TrDataSet = TrData_Set. } } public double CC_LearningRate { get { return _CC_LearningRate. _BP_Reports = BPReports.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000087 00000088 00000089 00000090 00000091 00000092 00000093 00000094 00000095 00000096 00000097 00000098 00000099 00000100 00000101 00000102 00000103 00000104 00000105 00000106 00000107 00000108 00000109 00000110 00000111 00000112 00000113 00000114 00000115 00000116 _BP_HiddenL = BPHiddenL. } } public int CC_MaxNeurons { get { return _CC_MaxNeurons. _BP_LearningRate = BPLearningRate. } } #endregion #region genetic_algorithms //Genetic Algorithm public int GA_PopSize { get { return _GA_PopSize. } } public int CC_Reports { get { return _CC_Reports.

} } public bool BP_ITerr { get { return _BP_ITerr. } } public int GA_Reports { get { return _GA_Reports. } } public int GA_HiddenL { get { return _GA_HiddenL. } } public int BP_HiddenL { get { return _BP_HiddenL. } } public double GA_Terr { get { return _GA_Terr.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000117 00000118 00000119 00000120 00000121 00000122 00000123 00000124 00000125 00000126 00000127 00000128 00000129 00000130 00000131 00000132 00000133 00000134 00000135 00000136 00000137 00000138 00000139 00000140 00000141 00000142 00000143 00000144 00000145 00000146 { get { return _GA_GenLimit. } } public int GA_Crossover { get { return _GA_Crossover. } } public int GA_Mutation { get { return _GA_Mutation. } } public bool GA_ITerr { get { return _GA_ITerr. } } #endregion Page 79 of 225 . } } public double BP_LearningRate { get { return _BP_LearningRate. } } #endregion #region back_propagation //Back propagation public int BP_EpochLimit { get { return _BP_EpochLimit. } } public int BP_Reports { get { return _BP_Reports. } } public double BP_Terr { get { return _BP_Terr.

“FrmMain. System. System. System. System.Data.Controller { public partial class FrmMain : Form Page 80 of 225 .2 .1. using _05025397.cs” 00000001 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 using using using using using using using using using System.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000147 00000148 00000149 00000150 00000151 00000152 00000153 00000154 00000155 00000156 00000157 } #endregion #region external //General public int TrDataSet { get { return _TrDataSet. namespace _05025397. System. } } #endregion } 9.Linq.Drawing.IO. System.Text. using ZedGraph. System.Generic.ComponentModel.Forms. System.Windows.Collections. } } public int TrainingAlgorithm { get { return trainingalg.1.

William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000018 00000019 00000020 00000021 00000022 00000023 00000024 00000025 00000026 00000027 00000028 00000029 00000030 00000031 00000032 00000033 00000034 00000035 00000036 00000037 00000038 00000039 00000040 00000041 00000042 00000043 00000044 00000045 00000046 00000047 { #region data /********************\ |* DATA *| \********************/ //The neuralnetwork Controller. this.RunWorkerCompleted += new RunWorkerCompletedEventHandler(bw_RunWorkerCompleted).Working workingdialog = new Controller. //Working dialog Controller. bw.Algorithms. #endregion #region constructor /********************\ |* CONSTRUCTOR *| \********************/ public FrmMain() { InitializeComponent(). //Background thread static BackgroundWorker bw = new BackgroundWorker().Icon = Properties. } #endregion #region internal_functions /********************\ Page 81 of 225 . bw.DoWork += new DoWorkEventHandler(bw_DoWork).icon.Resources.TrainingAlgorithm network.Working().

Clear().CurveList. MSEpane.AxisChange(). try { if (e. SymbolType.Text = "MSE".GraphPane.Legend. // Hide the legend MSEpane.TrainingAlgorithm == 0) { Page 82 of 225 . Color. GraphPane MSEpane = this.Title.zedGraphControl1.Text = "Mean Squared Error".YAxis.zedGraphControl1.Title.Title.NetworkUpdateEventArgs e) { LoadDataset(e. this.Red.GraphPane. //Titles MSEpane.IsVisible = false.zedGraphControl1. } private void updatenetwork(object sender.Circle).AddCurve("". Classes.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000048 00000049 00000050 00000051 00000052 00000053 00000054 00000055 00000056 00000057 00000058 00000059 00000060 00000061 00000062 00000063 00000064 00000065 00000066 00000067 00000068 00000069 00000070 00000071 00000072 00000073 00000074 00000075 00000076 00000077 |* INTERNAL *| \********************/ private void plotgraphsMSE(PointPairList MSE) { this.TrDataSet).XAxis. MSEpane. MSEpane.Text = "Iteration". MSE.

GA_PopSize. e.GA_Reports.CC_MaxNeurons. e. e.GA_GenLimit.CC_ITerr. txtTestData).Algorithms. txtTrainData. e. e.BP_HiddenL.CascadeCorrelation (e.BP_Terr.BP_LearningRate. } if (e. e. e.CC_Terr.Algorithms.Algorithms.CC_LearningRate.TrainingAlgorithm == 2) { network = new Controller. txtTestData). txtTestData).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000078 00000079 00000080 00000081 00000082 00000083 00000084 00000085 00000086 00000087 00000088 00000089 00000090 00000091 00000092 00000093 00000094 00000095 00000096 00000097 00000098 00000099 00000100 00000101 00000102 00000103 00000104 00000105 00000106 00000107 network = new Controller. e. txtTrainData.CC_Reports.BP_Reports.GA_Crossover.GA_ITerr.TrainingAlgorithm == 1) { network = new Controller. e.GA_Mutation.GeneticAlgorithm (e. e. e.BP_EpochLimit. e. txtTrainData.BP_ITerr.BackProp (e. e. e. e. } if (e.GA_HiddenL. } } catch { Page 83 of 225 .GA_Terr. e.

Wiz1().network_details().NetworkUpdateHandler(updatenetwork).Exit().Text = Properties. MessageBoxButtons.Resources.Show("Error creating new network".Wiz1 wizard = new Controller.XORtrain.Text = network.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000108 00000109 00000110 00000111 00000112 00000113 00000114 00000115 00000116 00000117 00000118 00000119 00000120 00000121 00000122 00000123 00000124 00000125 00000126 00000127 00000128 00000129 00000130 00000131 00000132 00000133 00000134 00000135 00000136 00000137 MessageBox. Page 84 of 225 . wizard. "Error". Application.Resources.Wiz1.Text = Properties.NetworkUpdated += new Controller. wizard.OK. } private void LoadDataset(int Data_set) { switch (Data_set) { case (int)Dataset.XOR: txtTrainData. txtTestData. } if (network != null) { txtNetSettings.Error). MessageBoxIcon.XORtest. } } private void new_network() { Controller.ShowDialog().

break. string test = "". } Page 85 of 225 . break.FISHERtest.Text = Properties.CUSTOM: LoadCustomDataSet(). } } private void LoadCustomDataSet() { string trainpath = SelectTextFile(".Resources. MessageBoxIcon.Show("Error: Problem loading training data". } catch { MessageBox. string testpath = SelectTextFile(".\\".Resources. case (int)Dataset.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000138 00000139 00000140 00000141 00000142 00000143 00000144 00000145 00000146 00000147 00000148 00000149 00000150 00000151 00000152 00000153 00000154 00000155 00000156 00000157 00000158 00000159 00000160 00000161 00000162 00000163 00000164 00000165 00000166 00000167 break.OK.\\".FISHER: txtTrainData. txtTestData.Error). "Select Testing Data"). string train = "". "Select Training Data"). try { train = LoadTxtFile(trainpath). MessageBoxButtons. case (int)Dataset.FISHERtrain.Text = Properties. "Error!".

Text = test. MessageBoxIcon. } private string LoadTxtFile(string path) { string data = "".OK. txtTrainData.Open. MessageBoxButtons.Show("Error reading selected file: " + path. Reader = new StreamReader(File). "Error!". MessageBoxIcon. StreamReader Reader = null. try { File = new FileStream(path. "Error!". MessageBoxButtons. Page 86 of 225 .Error).ReadToEnd(). FileMode.Read). data = Reader. } catch { MessageBox.OK. } catch { MessageBox.Show("Error: Problem loading testing data". FileStream File = null. FileAccess.Text = train. } txtTestData.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000168 00000169 00000170 00000171 00000172 00000173 00000174 00000175 00000176 00000177 00000178 00000179 00000180 00000181 00000182 00000183 00000184 00000185 00000186 00000187 00000188 00000189 00000190 00000191 00000192 00000193 00000194 00000195 00000196 00000197 try { test = LoadTxtFile(testpath).Error).

*". } #endregion #region menu_items /********************\ |* MENU ITEMS *| \********************/ Page 87 of 225 .Close(). } finally { if (Reader != null) Reader. if (File != null) File.ShowDialog() == DialogResult.Title = title. string title) { OpenFileDialog dialog = new OpenFileDialog().*)|*.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000198 00000199 00000200 00000201 00000202 00000203 00000204 00000205 00000206 00000207 00000208 00000209 00000210 00000211 00000212 00000213 00000214 00000215 00000216 00000217 00000218 00000219 00000220 00000221 00000222 00000223 00000224 00000225 00000226 00000227 data = "". } private string SelectTextFile(string initialDirectory.txt)|*.Filter = "txt files (*. } return data.txt|All files (*.OK) ? dialog.Close(). dialog. return (dialog. dialog.InitialDirectory = initialDirectory.FileName : null. dialog.

EventArgs e) { new_network(). EventArgs e) { if (network != null) { if (network. } private void exitToolStripMenuItem_Click(object sender. } } } private void aboutToolStripMenuItem1_Click(object sender.Trained != false) { network.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000228 00000229 00000230 00000231 00000232 00000233 00000234 00000235 00000236 00000237 00000238 00000239 00000240 00000241 00000242 00000243 00000244 00000245 00000246 00000247 00000248 00000249 00000250 00000251 00000252 00000253 00000254 00000255 00000256 00000257 private void newNetworkToolStripMenuItem_Click(object sender. EventArgs e) { FrmAbout About = new FrmAbout().Test(). } private void runToolStripMenuItem_Click(object sender.Close().Show(). EventArgs e) { this. } #endregion #region buttons Page 88 of 225 . About.

if (!bw. EventArgs e) { if (network != null) { if (network. EventArgs e) { //Shouldn't be busy if the user managed to click this //but just make 100% certain.Trained) { bw. } } } } private void btnTest_Click(object sender. EventArgs e) { new_network().Trained) Page 89 of 225 .IsBusy) { if (network != null) { if (!network.RunWorkerAsync().ShowDialog(). } private void btnTrain_Click(object sender.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000258 00000259 00000260 00000261 00000262 00000263 00000264 00000265 00000266 00000267 00000268 00000269 00000270 00000271 00000272 00000273 00000274 00000275 00000276 00000277 00000278 00000279 00000280 00000281 00000282 00000283 00000284 00000285 00000286 00000287 /********************\ |* BUTTONS *| \********************/ private void btnNew_Click(object sender. workingdialog.

William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000288 00000289 00000290 00000291 00000292 00000293 00000294 00000295 00000296 00000297 00000298 00000299 00000300 00000301 00000302 00000303 00000304 00000305 00000306 00000307 00000308 00000309 00000310 00000311 00000312 00000313 00000314 00000315 00000316 00000317 { if (!network.Tested) { bool success = network.Train(). txtOutput. DoWorkEventArgs e) { bool success = true. } } } } } #endregion #region bg_worker /********************\ |* BG WORKER *| \********************/ private void bw_DoWork(object sender.Text.ScrollToCaret().Refresh(). txtOutput.Test().Length. if (network != null) { success = network.ReportData. } Page 90 of 225 .Clear().Text = network.SelectionStart = txtOutput. if (success) { txtOutput. txtOutput. txtOutput.

if (success) { txtOutput. RunWorkerCompletedEventArgs e) { bool success. } else { MessageBox. //Check for errors if (e. MessageBoxIcon.Clear(). } Page 91 of 225 .Error.ReportData.OK. txtOutput.Error).Text.Show("Error during network train: " + e.ScrollToCaret(). MessageBoxButtons.SelectionStart = txtOutput.Show("Error training network!".Error != null) { MessageBox.Length.Result = success. txtOutput.OK.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000318 00000319 00000320 00000321 00000322 00000323 00000324 00000325 00000326 00000327 00000328 00000329 00000330 00000331 00000332 00000333 00000334 00000335 00000336 00000337 00000338 00000339 00000340 00000341 00000342 00000343 00000344 00000345 00000346 00000347 e. plotgraphsMSE(network. txtOutput.Result. MessageBoxIcon.Error). "Error". "Error training network!". } private void bw_RunWorkerCompleted(object sender.Text = network.ToString().getGraphData()). } success = (bool) e. MessageBoxButtons.

3 . Page 92 of 225 . } #endregion } 9.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000348 00000349 00000350 00000351 00000352 00000353 00000354 00000355 00000356 00000357 00000358 00000359 00000360 00000361 00000362 00000363 00000364 00000365 00000366 } //Remove the dialog locking out the main form //and showing that we're currently working workingdialog. System.Linq.Collections.Drawing.1.1. System. System.“Wiz1.Generic. System.Hide(). } #endregion #region misc /********************\ |* MISC *| \********************/ private void FrmMain_Shown(object sender.ComponentModel.cs” 00000001 00000002 00000003 00000004 00000005 00000006 using using using using using using System. System.Data. EventArgs e) { //Run the new neural net wizard new_network().

Page 93 of 225 . private int algorithm. Classes. namespace _05025397. #endregion public partial class Wiz1 : Form { #region data /********************\ |* DATA *| \********************/ public delegate void NetworkUpdateHandler(object sender. public event NetworkUpdateHandler NetworkUpdated. enum Dataset { XOR. using _05025397. CC }.Controller { #region enums /********************\ |* ENUMS *| \********************/ enum Training { GA.Text. FISHER. BP.NetworkUpdateEventArgs e). private int Trainset. using System. CUSTOM }.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 00000018 00000019 00000020 00000021 00000022 00000023 00000024 00000025 00000026 00000027 00000028 00000029 00000030 00000031 00000032 00000033 00000034 00000035 00000036 using System.Windows.Forms.

Trainset = (int)Dataset. this.GA.XOR.Resources.Close().Icon = Properties.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000037 00000038 00000039 00000040 00000041 00000042 00000043 00000044 00000045 00000046 00000047 00000048 00000049 00000050 00000051 00000052 00000053 00000054 00000055 00000056 00000057 00000058 00000059 00000060 00000061 00000062 00000063 00000064 00000065 00000066 #endregion data #region constructor /********************\ |* CONSTRUCTOR *| \********************/ public Wiz1() { InitializeComponent(). } private void rBtnBackprop_CheckedChanged(object sender.icon. algorithm = (int)Training. EventArgs e) { this. EventArgs e) Page 94 of 225 . } #endregion #region control_events /********************\ |* CONTROL EVENTS *| \********************/ #region buttons /********************\ |* BUTTONS *| \********************/ private void btnCancel_Click(object sender.

Visible = true.Resources. lblCC.Visible = false. panCC. panGeneticAlgorithm.Resources.Visible = false.IMAGECascadeCorrelation.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000067 00000068 00000069 00000070 00000071 00000072 00000073 00000074 00000075 00000076 00000077 00000078 00000079 00000080 00000081 00000082 00000083 00000084 00000085 00000086 00000087 00000088 00000089 00000090 00000091 00000092 00000093 00000094 00000095 00000096 { if (rBtnBackprop.Visible = false.BP. algorithm = (int)Training. lblGA. lblCC.Checked == true) { picAlg.CC.Visible = false.Visible = false. algorithm = (int)Training.Visible = false. lblBP.Checked == true) { picAlg.Visible = true. panGeneticAlgorithm. panCC. panBackprop. lblGA. } } private void rBtnCascadeCorrelation_CheckedChanged(object sender.Image = _05025397.Properties.Visible = false.Visible = false. EventArgs e) { if (rBtnCascadeCorrelation.Properties. lblBP.IMAGEBackProp.Visible = true. Page 95 of 225 .Visible = true. panBackprop.Image = _05025397.

Visible = true.Visible = true. lblCC.Resources. } } private void btnNext_Click(object sender. EventArgs e) { if (rBtnGeneticAlgorithm.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000097 00000098 00000099 00000100 00000101 00000102 00000103 00000104 00000105 00000106 00000107 00000108 00000109 00000110 00000111 00000112 00000113 00000114 00000115 00000116 00000117 00000118 00000119 00000120 00000121 00000122 00000123 00000124 00000125 00000126 } } private void rBtnGeneticAlgorithm_CheckedChanged(object sender.Checked == true) { picAlg.Image = _05025397. lblBP.1).IMAGEGeneticAlgorithm. panGeneticAlgorithm. } } private void btnBack_Click(object sender. lblGA. algorithm = (int)Training. panCC. EventArgs e) { if (tabControl1. panBackprop.Visible = false.GA.Properties.SelectedIndex . EventArgs e) { Page 96 of 225 .SelectedIndex > 0) { tabControl1.Visible = false.Visible = false.Visible = false.SelectTab(tabControl1.

CC_ITerr.Text. txtGA_Terr. check. txtCC_TargetError. } else { VerifyData check = new VerifyData (algorithm.Text. CC_ITerr. txtBP_EpochLimit. txtCC_MaxNeurons.TabCount . check.CC_Reports.Checked. check. txtBP_LearnRate. txtGA_MaxGen. Page 97 of 225 .Verified) { Classes. check.Text.NetworkUpdateEventArgs (check. txtGA_ProbX.Text.CC_LearningRate.Text.SelectedIndex < tabControl1.Text.1) { tabControl1. check.BP_EpochLimit. chkGA_IgnoreTarget. check. txtGA_HiddenNodes.GA_GenLimit. txtCC_LearningRate.Checked.TrainingAlgorithm.Text. txtGA_PopSize. Trainset).GA_Crossover.Text.BP_HiddenL. txtCC_Report.Text.Text.GA_Terr.Text.Text. check. check.GA_Reports. check.GA_HiddenL.Text.SelectTab(tabControl1.GA_Mutation.GA_PopSize. txtGA_Report.CC_Terr.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000127 00000128 00000129 00000130 00000131 00000132 00000133 00000134 00000135 00000136 00000137 00000138 00000139 00000140 00000141 00000142 00000143 00000144 00000145 00000146 00000147 00000148 00000149 00000150 00000151 00000152 00000153 00000154 00000155 00000156 if (tabControl1.GA_ITerr. txtBP_HiddenNodes.Text.NetworkUpdateEventArgs args = new Classes. txtGA_ProbM.Text. check. chkBP_IgnoreTarget. if (check.Text. check. txtBP_Report. txtBP_Terr. check. check.CC_MaxNeurons. check. check.Checked.SelectedIndex + 1).

MessageBoxButtons.TrDataSet).BP_Terr. EventArgs e) { if (chkDatasetFisher.Show("Error Checking Data: Data could not be verified". check. MessageBoxIcon.Resources.OK. check.Image = _05025397. args). Trainset = (int)Dataset. } } } #endregion #region checkboxes /********************\ |* CHECKBOXES *| \********************/ private void chkDatasetFisher_CheckedChanged(object sender. "Error Checking Data".IMAGEIrisFlowers.BP_LearningRate.Properties.BP_ITerr. } else { MessageBox. NetworkUpdated(this.Checked) { picDataset.Exclamation).BP_Reports.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000157 00000158 00000159 00000160 00000161 00000162 00000163 00000164 00000165 00000166 00000167 00000168 00000169 00000170 00000171 00000172 00000173 00000174 00000175 00000176 00000177 00000178 00000179 00000180 00000181 00000182 00000183 00000184 00000185 00000186 check.FISHER. this.Dispose(). check. check. EventArgs e) Page 98 of 225 . } } private void chkDatasetXOR_CheckedChanged(object sender.

Visible == true) { txtCC_TargetError.CUSTOM. EventArgs e) { if (tabControl1.IMAGEXOR. } } private void chkDatasetCustom_CheckedChanged(object sender. EventArgs e) { if (chkDatasetCustom.Resources. } } #endregion #region misc /********************\ |* MISC *| \********************/ private void tabControl1_SelectedIndexChanged(object sender. Trainset = (int)Dataset.XOR.SelectedIndex == 1) { if (panCC.Focus(). Trainset = (int)Dataset.Properties.Resources.IMAGEQuestionMark.Checked) { picDataset.Image = _05025397.Checked) { picDataset.Properties.Image = _05025397. } Page 99 of 225 .William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000187 00000188 00000189 00000190 00000191 00000192 00000193 00000194 00000195 00000196 00000197 00000198 00000199 00000200 00000201 00000202 00000203 00000204 00000205 00000206 00000207 00000208 00000209 00000210 00000211 00000212 00000213 00000214 00000215 00000216 { if (chkDatasetXOR.

Focus(). EventArgs e) { this.icon. tabControl1.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000217 00000218 00000219 00000220 00000221 00000222 00000223 00000224 00000225 00000226 00000227 00000228 00000229 00000230 00000231 00000232 00000233 00000234 00000235 00000236 } if (panBackprop.Visible == true) { txtGA_PopSize. } #endregion #endregion } Page 100 of 225 .Focus().Icon = Properties. } } } private void Wiz1_Load(object sender.SelectTab(0).Resources. } if (panGeneticAlgorithm.Visible == true) { txtBP_EpochLimit.

System.“FrmAbout.Windows. System. namespace _05025397.Linq. System.Drawing.Generic. } #endregion #region form /********************\ |* FORM *| \********************/ private void FrmAbout_Load(object sender.ComponentModel.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 9.1.Controller { public partial class FrmAbout : Form { #region constructor /********************\ |* CONSTRUCTOR *| \********************/ public FrmAbout() { InitializeComponent(). EventArgs e) { Page 101 of 225 . System. System.Forms.cs” 00000001 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 00000018 00000019 00000020 00000021 00000022 00000023 00000024 00000025 00000026 00000027 00000028 00000029 using using using using using using using using System.4 . System. System.Text.Data.1.Collections.

OK. System.Resources. MessageBoxIcon.LICENCE_LGPL21. textBox5.Icon = Properties. } #endregion #region linklabels /********************\ |* LINKLABELS *| \********************/ private void linkLabel1_LinkClicked(object sender. LinkLabelLinkClickedEventArgs e) Page 102 of 225 . "Undefined Application".Text). try { // Navigate to a URL.icon.Diagnostics. this.Text = Application.LinkVisited = true.Text = Properties.Show("No e-mail program defined.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000030 00000031 00000032 00000033 00000034 00000035 00000036 00000037 00000038 00000039 00000040 00000041 00000042 00000043 00000044 00000045 00000046 00000047 00000048 00000049 00000050 00000051 00000052 00000053 00000054 00000055 00000056 00000057 00000058 00000059 this.Process.Resources. label5. } } private void linkLabel4_LinkClicked(object sender. LinkLabelLinkClickedEventArgs e) { // Specify that the link was visited.Information). } catch { MessageBox.ProductVersion.". MessageBoxButtons.linkLabel1.Start("mailto:" + linkLabel1.

Show("Error opening link.Text).Start (linkLabel5. "Error".Process. } catch { MessageBox.". MessageBoxIcon. "Error". MessageBoxIcon. LinkLabelLinkClickedEventArgs e) { try { System.".Process.Text).Show("Error opening link.Error). LinkLabelLinkClickedEventArgs e) Page 103 of 225 .Error). } } private void linkLabel5_LinkClicked(object sender.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000060 00000061 00000062 00000063 00000064 00000065 00000066 00000067 00000068 00000069 00000070 00000071 00000072 00000073 00000074 00000075 00000076 00000077 00000078 00000079 00000080 00000081 00000082 00000083 00000084 00000085 00000086 00000087 00000088 00000089 { try { System. MessageBoxButtons.Start (linkLabel4.Diagnostics. } catch { MessageBox. MessageBoxButtons.Diagnostics. } } private void linkLabel6_LinkClicked(object sender.OK.OK.

Show("Error opening link.".OK. } catch { MessageBox.Error).Process.Diagnostics.Start (linkLabel6.uk").snikks. MessageBoxIcon.Error). } } private void linkLabel2_LinkClicked(object sender. } } private void linkLabel7_LinkClicked(object sender.co. MessageBoxIcon. MessageBoxButtons.OK.Text). } catch { MessageBox.Process. "Error".Start ("http://www. LinkLabelLinkClickedEventArgs e) Page 104 of 225 .Show("Error opening link.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000090 00000091 00000092 00000093 00000094 00000095 00000096 00000097 00000098 00000099 00000100 00000101 00000102 00000103 00000104 00000105 00000106 00000107 00000108 00000109 00000110 00000111 00000112 00000113 00000114 00000115 00000116 00000117 00000118 00000119 { try { System.".Diagnostics. MessageBoxButtons. LinkLabelLinkClickedEventArgs e) { try { System. "Error".

groupBox2.".Visible = false.Start ("http://www.Show("Error opening link.1.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000120 00000121 00000122 00000123 00000124 00000125 00000126 00000127 00000128 00000129 00000130 00000131 00000132 00000133 00000134 00000135 00000136 00000137 00000138 00000139 00000140 00000141 00000142 00000143 00000144 00000145 00000146 00000147 00000148 00000149 { try { System.Process.gnu.Visible = true. EventArgs e) { if (tabControl1.Visible = false.Visible = false. MessageBoxIcon. } catch { MessageBox. } } #endregion #region buttons /********************\ |* BUTTONS *| \********************/ private void btnDatasets_Click(object sender.Error). pictureBox1.Visible != true) { groupBox1.OK.org/licenses/old-licenses/lgpl-2. } else Page 105 of 225 .Diagnostics. MessageBoxButtons. "Error". tabControl1.html").

} else { groupBox1. EventArgs e) { this.Visible = false.Visible != true) { groupBox2. tabControl1. false. groupBox1.Visible } } private void btnLicences_Click(object sender. } true. } } private void btnOkay_Click(object sender. EventArgs e) { if (groupBox2. groupBox2.Visible = false.Visible = false.Visible = false. = false. pictureBox1.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000150 00000151 00000152 00000153 00000154 00000155 00000156 00000157 00000158 00000159 00000160 00000161 00000162 00000163 00000164 00000165 00000166 00000167 00000168 00000169 00000170 00000171 00000172 00000173 00000174 00000175 00000176 00000177 00000178 00000179 { groupBox1.Visible = false.Visible = true.Visible pictureBox1.Visible = tabControl1. tabControl1.Dispose().Visible = groupBox2.Visible = true. pictureBox1. Page 106 of 225 . = true.Visible = true.

Checked == true) textBox5. EventArgs e) { if (radioButton3.LICENCE_LGPL30. } #endregion } Page 107 of 225 .Text = Properties.Resources.LICENCE_GPL30.Text = Properties. EventArgs e) { if (radioButton2.Resources.Checked == true) textBox5. } private void radioButton3_CheckedChanged(object sender. } private void radioButton2_CheckedChanged(object sender.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000180 00000181 00000182 00000183 00000184 00000185 00000186 00000187 00000188 00000189 00000190 00000191 00000192 00000193 00000194 00000195 00000196 00000197 00000198 00000199 00000200 00000201 00000202 00000203 00000204 00000205 } #endregion #region radiobuttons /********************\ |* RADIOBUTTONS *| \********************/ private void radioButton1_CheckedChanged(object sender.Checked == true) textBox5.Resources. EventArgs e) { if (radioButton1.Text = Properties.LICENCE_LGPL21.

Linq. namespace _05025397. System.Play().Beep.1.Collections. System.Icon = Properties. System.Controller { public partial class Working : Form { #region functions /********************\ |* FUNCTIONS *| \********************/ public Working() { InitializeComponent().icon. System. System.Data.5 .ComponentModel.Resources.“Working.Forms.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 9.1.Windows. } private void Working_Click(object sender. System.cs” 00000001 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 00000018 00000019 00000020 00000021 00000022 00000023 00000024 00000025 00000026 00000027 00000028 00000029 using using using using using using using using System. System.Drawing.SystemSounds. this.Text. } #endregion } Page 108 of 225 .Media.Generic. EventArgs e) { System.

Collections.Forms. System.Text.1 .Linq. System. namespace _05025397. System.2.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000030 } 9. //Dataset identifer //0 = XOR Page 109 of 225 .2 . //Training Algorithm Identifier //1 = genetic algorith //2 = backprop //3 = cascade correlation private int trainingalg.Controller { public class VerifyData { #region data /********************\ |* DATA *| \********************/ //Dataverified variable private bool verified.Windows.1.“VerifyData.Controller 9.cs” 00000001 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 00000018 00000019 00000020 00000021 00000022 00000023 00000024 00000025 using using using using using System.1.Generic. System.

private int _GA_Reports. #endregion Page 110 of 225 . private int _BP_Reports. private double _GA_Terr. private double _BP_LearningRate. private int _GA_GenLimit. //Genetic Algorithm private int _GA_PopSize. private double _BP_Terr. private int _GA_HiddenL. private int _CC_Reports. private bool _CC_ITerr. private bool _GA_ITerr.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000026 00000027 00000028 00000029 00000030 00000031 00000032 00000033 00000034 00000035 00000036 00000037 00000038 00000039 00000040 00000041 00000042 00000043 00000044 00000045 00000046 00000047 00000048 00000049 00000050 00000051 00000052 00000053 00000054 00000055 //1 = FISHERS IRIS DATA //2 = CUSTOM private int _TDataSet. private int _GA_Crossover. private bool _BP_ITerr. //BackProp private int _BP_EpochLimit. private double _CC_LearningRate. private int _BP_HiddenL. private int _CC_MaxNeurons. //Cascade Correlation private double _CC_Terr. private int _GA_Mutation.

XOR && _TDataSet != Page 111 of 225 . if (_TDataSet != (int)Dataset. string BPReports. string GACrossOver. string GAPopSize. bool CCITerr. string BPTerr. string BPEpochLimit. //DataSet _TDataSet = TData_Set. string BPLearningRate. string CCReports. string GAReports. string GAHiddenL. will be changed to //false if any data verification errors occur verified = true. bool BPITerr. string CCMaxNeurons. string GATerr. bool GAITerr. //Algorithm trainingalg = Algorithm. string CCTerr. int TData_Set) { //Initially set to true. string BPHiddenL. string GAGenLimit.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000056 00000057 00000058 00000059 00000060 00000061 00000062 00000063 00000064 00000065 00000066 00000067 00000068 00000069 00000070 00000071 00000072 00000073 00000074 00000075 00000076 00000077 00000078 00000079 00000080 00000081 00000082 00000083 00000084 00000085 #region constructor /********************\ |* CONSTRUCTOR *| \********************/ public VerifyData (int Algorithm. string CCLearningRate. string GAMutation.

_CC_MaxNeurons = Convert.ToInt32(CCMaxNeurons). } } } Page 112 of 225 .Error).ToInt32(CCReports). _CC_ITerr = CCITerr.OK.ToSingle(CCLearningRate). _CC_Reports = Convert. "Parsing Error".ToSingle(CCTerr).CUSTOM) verified = false.Show ("Parsing Error: Please check supplied cascade" + "correlation data values". if (verified) { if (trainingalg == 2) { try { //Cascade correlation _CC_Terr = Convert. } catch (FormatException) { MessageBox.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000086 00000087 00000088 00000089 00000090 00000091 00000092 00000093 00000094 00000095 00000096 00000097 00000098 00000099 00000100 00000101 00000102 00000103 00000104 00000105 00000106 00000107 00000108 00000109 00000110 00000111 00000112 00000113 00000114 00000115 (int)Dataset. MessageBoxIcon.FISHER && _TDataSet != (int)Dataset. verified = false. _CC_LearningRate = Convert. MessageBoxButtons.

verified = false. } catch (FormatException) { MessageBox.ToInt32(GAReports). _GA_Crossover = Convert. MessageBoxButtons. _GA_GenLimit = Convert.ToInt32(GAPopSize).ToInt32(GAHiddenL).OK.ToInt32(GAMutation).ToInt32(GAGenLimit). "Parsing Error". _GA_Reports = Convert. _GA_ITerr = GAITerr. MessageBoxIcon.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000116 00000117 00000118 00000119 00000120 00000121 00000122 00000123 00000124 00000125 00000126 00000127 00000128 00000129 00000130 00000131 00000132 00000133 00000134 00000135 00000136 00000137 00000138 00000139 00000140 00000141 00000142 00000143 00000144 00000145 if (verified) { if (trainingalg == 0) { try { //Genetic algorithm _GA_PopSize = Convert. } } } if (verified) Page 113 of 225 . _GA_HiddenL = Convert. _GA_Mutation = Convert.Show ("Parsing Error: Please check supplied" +" genetic algorithm data values".ToInt32(GACrossOver).ToSingle(GATerr). _GA_Terr = Convert.Error).

_BP_Reports = Convert. } Page 114 of 225 .William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000146 00000147 00000148 00000149 00000150 00000151 00000152 00000153 00000154 00000155 00000156 00000157 00000158 00000159 00000160 00000161 00000162 00000163 00000164 00000165 00000166 00000167 00000168 00000169 00000170 00000171 00000172 00000173 00000174 00000175 { if (trainingalg == 1) { try { //Back-propagation _BP_EpochLimit = Convert. _BP_LearningRate = Convert. "Parsing Error". _BP_Terr = Convert.ToInt32(BPHiddenL).ToDouble(BPTerr). verified = false. MessageBoxIcon.ToInt32(BPReports). _BP_HiddenL = Convert. _BP_ITerr = BPITerr. } catch (FormatException) { MessageBox. MessageBoxButtons.Show ("Parsing Error: Please check " + "supplied back propagation data values".OK. } } } if (verified) { verified = checkforsanity().Error).ToInt32(BPEpochLimit).ToDouble(BPLearningRate).

} if (_GA_Terr < 0 || _GA_Terr > 1) sanity = false. if (_GA_PopSize < 3) sanity = false. Page 115 of 225 . } if (_GA_Mutation < 0 || _GA_Mutation > 100) { sanity = false. if (_GA_Crossover < 0 || _GA_Crossover > 100) { sanity = false.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000176 00000177 00000178 00000179 00000180 00000181 00000182 00000183 00000184 00000185 00000186 00000187 00000188 00000189 00000190 00000191 00000192 00000193 00000194 00000195 00000196 00000197 00000198 00000199 00000200 00000201 00000202 00000203 00000204 00000205 } #endregion #region internal /********************\ |* INTERNAL *| \********************/ private bool checkGAsanity() { bool sanity = true. if (_GA_GenLimit < 1) sanity = false.

if (_BP_Reports < 1 || _BP_Reports > _BP_EpochLimit) sanity = false.OK.Error). MessageBoxButtons.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000206 00000207 00000208 00000209 00000210 00000211 00000212 00000213 00000214 00000215 00000216 00000217 00000218 00000219 00000220 00000221 00000222 00000223 00000224 00000225 00000226 00000227 00000228 00000229 00000230 00000231 00000232 00000233 00000234 00000235 if (_GA_Reports < 1 || _GA_Reports > _GA_GenLimit) sanity = false.Show ("Sanity Check Fail: Please check supplied " + "genetic algorithm data values". if (_GA_HiddenL < 1) sanity = false. MessageBoxIcon. return sanity. if (_BP_HiddenL < 1) sanity = false. "Sanity Check Fail". Page 116 of 225 . if (!sanity) MessageBox. if (_BP_EpochLimit < 1) sanity = false. } private bool checkBPsanity() { bool sanity = true.

MessageBoxIcon.Show ("Sanity Check Fail: Please check supplied " + "back propagation data values". "Sanity Check Fail". if (_CC_Terr < 0 || _CC_Terr > 1) sanity = false. if (!sanity) MessageBox.Error).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000236 00000237 00000238 00000239 00000240 00000241 00000242 00000243 00000244 00000245 00000246 00000247 00000248 00000249 00000250 00000251 00000252 00000253 00000254 00000255 00000256 00000257 00000258 00000259 00000260 00000261 00000262 00000263 00000264 00000265 if (_BP_Terr < 0 || _BP_Terr > 1) sanity = false. MessageBoxButtons. if (_CC_MaxNeurons < 1) sanity = false. } private bool checkCCsanity() { bool sanity = true. if (_BP_LearningRate < 0 || _BP_LearningRate > 1) sanity = false. return sanity. Page 117 of 225 . if (_CC_Reports < 1 || _CC_Reports > _CC_MaxNeurons) sanity = false.OK.

// // // // // bool Basically we're making sure the data is clean before it goes anywhere near a DLL.Error). } private bool checkforsanity() { //Some of the values that are allowed through here would not produce //good results . "Sanity Check Fail". only //halt any values from proceeding that are so wrong they could cause //serious errors later down the line.Show ("Sanity Check Fail: Please check supplied " + "cascade correlation data values". if they're greater than 1 but less than 100 scale them appropriately.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000266 00000267 00000268 00000269 00000270 00000271 00000272 00000273 00000274 00000275 00000276 00000277 00000278 00000279 00000280 00000281 00000282 00000283 00000284 00000285 00000286 00000287 00000288 00000289 00000290 00000291 00000292 00000293 00000294 00000295 if (_CC_LearningRate < 0 || _CC_LearningRate > 1) sanity = false. sanity = true. MessageBoxIcon. return sanity. We're also doing some very minor processing of the probability values. MessageBoxButtons. if (!sanity) MessageBox. if (trainingalg < 0 || trainingalg > 2) Page 118 of 225 .however I don't want to prevent any flexibility.OK.

Show ("Sanity Check Fail: Training Algorithm not " + "selected or selected incorrectly". } if (trainingalg == 2) { sanity = checkCCsanity(). MessageBoxIcon.OK. MessageBoxButtons. } Page 119 of 225 . } if (sanity) { if (trainingalg == 0) { sanity = checkGAsanity().William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000296 00000297 00000298 00000299 00000300 00000301 00000302 00000303 00000304 00000305 00000306 00000307 00000308 00000309 00000310 00000311 00000312 00000313 00000314 00000315 00000316 00000317 00000318 00000319 00000320 00000321 00000322 00000323 00000324 00000325 { sanity = false. } } return sanity. "Sanity Check Fail". } if (trainingalg == 1) { sanity = checkBPsanity(). MessageBox.Error).

} } public int CC_MaxNeurons { get { return _CC_MaxNeurons. } } public int CC_Reports { get { return _CC_Reports. } } public int GA_GenLimit { get { return _GA_GenLimit. } } #endregion #region genetic_algorithm //Genetic Algorithm public int GA_PopSize { get { return _GA_PopSize. } } public bool CC_ITerr { get { return _CC_ITerr. } } public double CC_LearningRate { get { return _CC_LearningRate.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000326 00000327 00000328 00000329 00000330 00000331 00000332 00000333 00000334 00000335 00000336 00000337 00000338 00000339 00000340 00000341 00000342 00000343 00000344 00000345 00000346 00000347 00000348 00000349 00000350 00000351 00000352 00000353 00000354 00000355 #endregion #region getters/setters /********************\ |* EXTERNAL *| \********************/ #region cascade_correlation //Cascade Correlation public double CC_Terr { get { return _CC_Terr. } } public int GA_Mutation Page 120 of 225 . } } public int GA_Crossover { get { return _GA_Crossover.

} } _GA_Reports. } } public bool BP_ITerr { get { return _BP_ITerr. } } public int BP_HiddenL { get { return _BP_HiddenL. } } _GA_ITerr. } } _GA_Terr. } } _GA_HiddenL. } } public double BP_Terr { get { return _BP_Terr. } } #endregion #endregion #region external //Verification Page 121 of 225 . } } public int BP_Reports { get { return _BP_Reports.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000356 00000357 00000358 00000359 00000360 00000361 00000362 00000363 00000364 00000365 00000366 00000367 00000368 00000369 00000370 00000371 00000372 00000373 00000374 00000375 00000376 00000377 00000378 00000379 00000380 00000381 00000382 00000383 00000384 00000385 { get { return public double GA_Terr { get { return public int GA_Reports { get { return public int GA_HiddenL { get { return public bool GA_ITerr { get { return #endregion _GA_Mutation. } } public double BP_LearningRate { get { return _BP_LearningRate. } } #region back_propagation //Back propagation public int BP_EpochLimit { get { return _BP_EpochLimit.

} } #endregion } Page 122 of 225 . } } //TrainingAlgorithm public int TrainingAlgorithm { get { return trainingalg.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000386 00000387 00000388 00000389 00000390 00000391 00000392 00000393 00000394 00000395 00000396 00000397 00000398 } public bool Verified { get { return verified. } } //Dataset public int TrDataSet { get { return _TDataSet.

System.Windows.Text. private ArrayList trinputs.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 9. System. System.2 .Collections. private ArrayList teinputs. private ArrayList altest. private string[] data.Forms.1. System.2. private ArrayList troutputs.“DatasetParser. System. private string[] test. private ArrayList aldata.Linq.Controller { public class DatasetParser { #region data /********************\ |* DATA *| \********************/ private bool _Verified.Collections. Page 123 of 225 . namespace _05025397.cs” 00000001 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 00000018 00000019 00000020 00000021 00000022 00000023 00000024 00000025 00000026 00000027 00000028 00000029 using using using using using using System. private ArrayList teoutputs.Generic.

Replace("\r".Text = txtTest.Trim(). '\t'. '. ""). ""). aldata = new ArrayList().Text.Replace("\r". char[] delimiterChars = { '\n'. '.Text.Text = txtData. troutputs = new ArrayList(). trinputs = new ArrayList(). #endregion #region constructor /********************\ |* CONSTRUCTOR *| \********************/ public DatasetParser(RichTextBox txtData. altest = new ArrayList().'.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000030 00000031 00000032 00000033 00000034 00000035 00000036 00000037 00000038 00000039 00000040 00000041 00000042 00000043 00000044 00000045 00000046 00000047 00000048 00000049 00000050 00000051 00000052 00000053 00000054 00000055 00000056 00000057 00000058 00000059 //private Scaler scale.Text = txtTest. teinputs = new ArrayList().Trim(). txtTest. //Strip carriage returns and trim txtData. private double Noutputs. private double Niterations. private double Ninputs. ' '.' }. txtTest. teoutputs = new ArrayList().Text. '\r'. txtData.Text = txtData.Text. RichTextBox txtTest) { //Data init bool Verified = true. Page 124 of 225 .

number of inputs //and number of outputs Niterations = System. //Get the number of iterations. //parse the data into appropriately //structured arrays Verified = parseTrainingData(). try { for (int i = 3.Convert.Text.ToDouble(data[0]).ToDouble(data[1]).ToDouble(data[2]).Convert.Convert.Text. test = txtTest. } #endregion #region internal_functions /********************\ |* INTERNAL *| \********************/ private bool parseTrainingData() { bool success = true.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000060 00000061 00000062 00000063 00000064 00000065 00000066 00000067 00000068 00000069 00000070 00000071 00000072 00000073 00000074 00000075 00000076 00000077 00000078 00000079 00000080 00000081 00000082 00000083 00000084 00000085 00000086 00000087 00000088 00000089 //Split at the chars specified above data = txtData.Split(delimiterChars). _Verified = Verified.Split(delimiterChars). Ninputs = System. Verified = parseTestingData(). i < data. i++) Page 125 of 225 .Length. Noutputs = System.

RemoveAt(0). success = false. aldata. j < Noutputs.Add(aldata[0]). MessageBoxButtons. j++) { troutputs.RemoveAt(0).Count > 0) { for (int i = 0. aldata. } for (int j = 0. Page 126 of 225 . "Error".William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000090 00000091 00000092 00000093 00000094 00000095 00000096 00000097 00000098 00000099 00000100 00000101 00000102 00000103 00000104 00000105 00000106 00000107 00000108 00000109 00000110 00000111 00000112 00000113 00000114 00000115 00000116 00000117 00000118 00000119 { aldata. } private bool parseTestingData() { bool success = true.Add(aldata[0]).OK. } return success.Show("Error parsing training data". } while (aldata.Error).Convert. } } } catch { MessageBox.ToDouble(data[i])). i++) { trinputs. i < Ninputs.Add(System. MessageBoxIcon.

Add(System. "Error".Add(altest[0]).Show("Error parsing testing data". altest.Convert.Error).RemoveAt(0).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000120 00000121 00000122 00000123 00000124 00000125 00000126 00000127 00000128 00000129 00000130 00000131 00000132 00000133 00000134 00000135 00000136 00000137 00000138 00000139 00000140 00000141 00000142 00000143 00000144 00000145 00000146 00000147 00000148 00000149 try { for (int i = 3. i < Ninputs.Count > 0) { for (int i = 0. } Page 127 of 225 . j < Noutputs. success = false. i++) { teinputs.ToDouble(data[i])).Length. } return success.RemoveAt(0). } while (altest.OK. MessageBoxButtons. MessageBoxIcon. j++) { teoutputs. } for (int j = 0.Add(altest[0]). i++) { altest. i < data. } } } catch { MessageBox. altest.

Count / (int) Ninputs][]. troutputdata = new double[troutputs. k < Ninputs.Count / (int)Ninputs). l++) { for (int k = 0. i < (troutputs. i < (trinputs. for (int j = 0. for (int i = 0. l = 0. } public void GetTrainingData (out double[][] trinputdata. out double Nout) { Ninp = Ninputs. i++) trinputdata[i] = new double[(int) Ninputs]. out double Ninp.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000150 00000151 00000152 00000153 00000154 00000155 00000156 00000157 00000158 00000159 00000160 00000161 00000162 00000163 00000164 00000165 00000166 00000167 00000168 00000169 00000170 00000171 00000172 00000173 00000174 00000175 00000176 00000177 00000178 00000179 #endregion #region external_functions /********************\ |* EXTERNAL *| \********************/ public void GetStructure(out double iter.Count / (int) Noutputs][]. iter = Niterations. j+=(int) Ninputs.Count / (int)Noutputs). k++) { Page 128 of 225 . out double[][] troutputdata) { trinputdata = new double[trinputs. Nout = Noutputs.Count. j < trinputs. i++) troutputdata[i] = new double[(int) Noutputs]. for (int i = 0.

k < Ninputs.Count / (int)Ninputs][]. k++) { teinputdata[l][k] = (double)teinputs[j + k]. j += (int)Ninputs. out double[][] teoutputdata) { teinputdata = new double[teinputs. k++) { troutputdata[l][k] = (double)troutputs[j + k]. } } } public void GetTestingData (out double[][] teinputdata. Page 129 of 225 . i++) teoutputdata[i] = new double[(int)Noutputs].Count / (int)Noutputs). j < teinputs. i < (teoutputs. k < Noutputs.Count / (int)Noutputs][]. l++) { for (int k = 0.Count / (int)Ninputs). teoutputdata = new double[teoutputs.Count. j < troutputs.Count. j += (int)Noutputs. i++) teinputdata[i] = new double[(int)Ninputs]. l++) { for (int k = 0. for (int j = 0. i < (teinputs. l = 0. l = 0. for (int i = 0.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000180 00000181 00000182 00000183 00000184 00000185 00000186 00000187 00000188 00000189 00000190 00000191 00000192 00000193 00000194 00000195 00000196 00000197 00000198 00000199 00000200 00000201 00000202 00000203 00000204 00000205 00000206 00000207 00000208 00000209 trinputdata[l][k] = (double) trinputs[j + } } for (int j = 0. k]. for (int i = 0.

j < teoutputs.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000210 00000211 00000212 00000213 00000214 00000215 00000216 00000217 00000218 00000219 00000220 00000221 00000222 00000223 00000224 00000225 00000226 00000227 00000228 00000229 00000230 00000231 00000232 00000233 00000234 00000235 00000236 } } } for (int j = 0. Page 130 of 225 . j += (int)Noutputs.“Algorithms\TrainingAlgorithm. k < Noutputs. } } #endregion } 9.Count.3 . k++) { teoutputdata[l][k] = (double)teoutputs[j + k].2.cs” 00000001 using System. l++) { for (int k = 0. teoutputdata = teoutputs.1. } public bool Verified { get { return _Verified. l = 0. out ArrayList teoutputdata) { teinputdata = teinputs. } } } public void GetTestingDataAL (out ArrayList teinputdata.

Collections.Generic.Forms. protected int CurIt. System.Text. protected bool _ITerr.Linq. protected PointPairList BestErrList.Windows. private int _IterMax. namespace _05025397. Page 131 of 225 . ZedGraph.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 00000018 00000019 00000020 00000021 00000022 00000023 00000024 00000025 00000026 00000027 00000028 00000029 00000030 00000031 using using using using using using System. System. _05025397. private string _ReportData. System. protected bool _Tested. private int _ReportInterval.Controller. protected bool _Trained.Algorithms { public abstract class TrainingAlgorithm { #region data /********************\ |* DATA *| \********************/ private double _TargetError.

} #endregion Page 132 of 225 . #endregion #region constructor /********************\ |* CONSTRUCTOR *| \********************/ public TrainingAlgorithm(double Target_Error. if (_ITerr) _TargetError = -1. protected DatasetParser getdata. int Iter_Max. _ReportInterval = Reports. bool ITerr) { _TargetError = Target_Error. txtTest). _ITerr = ITerr. so the //network will loop to completion. getdata = new DatasetParser(txtData. int Reports.0 //this is impossible to reach. //If the ignore target error is true //then set the target error to -1. RichTextBox txtTest. BestErrList = new PointPairList(). _IterMax = Iter_Max. RichTextBox txtData.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000032 00000033 00000034 00000035 00000036 00000037 00000038 00000039 00000040 00000041 00000042 00000043 00000044 00000045 00000046 00000047 00000048 00000049 00000050 00000051 00000052 00000053 00000054 00000055 00000056 00000057 00000058 00000059 00000060 00000061 protected Random Rand = new Random().00.

public abstract bool Test().William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000062 00000063 00000064 00000065 00000066 00000067 00000068 00000069 00000070 00000071 00000072 00000073 00000074 00000075 00000076 00000077 00000078 00000079 00000080 00000081 00000082 00000083 00000084 00000085 00000086 00000087 00000088 00000089 00000090 00000091 #region internal /********************\ |* INTERNAL *| \********************/ protected void Report(string output) { _ReportData += output. } #endregion #region external /********************\ |* EXTERNAL *| \********************/ public abstract bool Train(). } get { return _TargetError. } public double TargetError { set { _TargetError = value. } } Page 133 of 225 . public PointPairList getGraphData() { return BestErrList. public abstract string network_details().

} } public string ReportData { get { return _ReportData. } } public bool Tested { get { return _Tested. } set { _Trained = value. } } public int IterMax { get { return _IterMax. } } public bool Trained { get { return _Trained. } } #endregion } Page 134 of 225 .William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000092 00000093 00000094 00000095 00000096 00000097 00000098 00000099 00000100 00000101 00000102 00000103 00000104 00000105 00000106 00000107 00000108 00000109 00000110 00000111 00000112 00000113 00000114 00000115 00000116 00000117 00000118 00000119 } public int ReportInterval { get { return _ReportInterval.

4 .cs” 00000001 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 00000018 00000019 00000020 00000021 00000022 00000023 00000024 00000025 00000026 00000027 00000028 00000029 using using using using using using using using using using System. System.Linq.Text. ZedGraph. private double _LearnRate. _05025397. namespace _05025397.1. #endregion #region constructor Page 135 of 225 .Forms. System.“Algorithms\CascadeCorrelation. System. private double _OutputL.Collections.2. System. FANN_Cascade CCNet.IO.Controller.Collections.Windows.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 9. FANN_Wrapper.Generic. System.Algorithms { public class CascadeCorrelation : TrainingAlgorithm { #region data /********************\ |* DATA *| \********************/ private double _InputL. System.

txtData. base.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000030 00000031 00000032 00000033 00000034 00000035 00000036 00000037 00000038 00000039 00000040 00000041 00000042 00000043 00000044 00000045 00000046 00000047 00000048 00000049 00000050 00000051 00000052 00000053 00000054 00000055 00000056 00000057 00000058 00000059 /********************\ |* CONSTRUCTOR *| \********************/ public CascadeCorrelation(double LearnRate. CCNet = new FANN_Cascade((int)_InputL.TargetError. double iterations. base. double TargetError. } #endregion #region internal_functions /********************\ |* INTERNAL *| \********************/ private void saveDllData(RichTextBox tr.IterMax).ReportInterval. Iter_Max. getdata.GetStructure(out iterations. txtTest. (int)_OutputL. int Iter_Max. base. _LearnRate. txtTest). RichTextBox te) { Page 136 of 225 . out _InputL. RichTextBox txtTest) : base(TargetError. RichTextBox txtData. bool ITerr. saveDllData(txtData. ITerr) { _LearnRate = LearnRate. out _OutputL). int Reports. Reports.

Close().Format ("Network Details: \n Input Layer: {0:d}".Close(). tw = new StreamWriter("dlltedata.dat"). output. (int)_OutputL)). tw.Append( string.Write(te. tw.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000060 00000061 00000062 00000063 00000064 00000065 00000066 00000067 00000068 00000069 00000070 00000071 00000072 00000073 00000074 00000075 00000076 00000077 00000078 00000079 00000080 00000081 00000082 00000083 00000084 00000085 00000086 00000087 00000088 00000089 TextWriter tw = new StreamWriter("dlltrdata. (int)_InputL)).Append( string. _LearnRate)). tw.Format ("\t\t\t\t\tOutput Layer: {0:d}".00000}".Append( string. output.Write(tr. tw. output. } #endregion #region external_functions /********************\ |* EXTERNAL *| \********************/ public override string network_details() { StringBuilder output = new StringBuilder().Text).Format ("\nLearning Rate: {0:0.Text).Append( Page 137 of 225 .dat"). output.

} catch { Page 138 of 225 .00000}".Append( string. base. Report(CCNet.Append( string. base.Format ("\t\t\t\t\t\tIgnore target error " + "(process all iterations): {0}".IterMax)). } public override bool Train() { bool success = true.Format ("\t\tMaximum Nodes: {0:d}"._tr_output). output.Append( string.ReportInterval)).Format ("\t\t\tTarget Error: {0:0.".William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000090 00000091 00000092 00000093 00000094 00000095 00000096 00000097 00000098 00000099 00000100 00000101 00000102 00000103 00000104 00000105 00000106 00000107 00000108 00000109 00000110 00000111 00000112 00000113 00000114 00000115 00000116 00000117 00000118 00000119 string. output.TargetError)).ToString(). _ITerr)).Format ("\nReport every {0:d} generations. base. try { success = CCNet. output.Train(). return output.

base._Trained = success. } success = getGraphDataFromDll().William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000120 00000121 00000122 00000123 00000124 00000125 00000126 00000127 00000128 00000129 00000130 00000131 00000132 00000133 00000134 00000135 00000136 00000137 00000138 00000139 00000140 00000141 00000142 00000143 00000144 00000145 00000146 00000147 00000148 00000149 MessageBox. "Error".Error).Show("Error running Cascade training!". MessageBoxButtons. } public bool getGraphDataFromDll() { bool success = true. MessageBoxIcon. for (int i = 0. try { double[] besterrlist = CCNet. success = false.OK. MessageBoxButtons.Error).OK.Add((double)i. "Error". success = false. return success. i < besterrlist.Show("Error getting graph data from DLL".Length. MessageBoxIcon. besterrlist[i]).besterrorlist. } Page 139 of 225 . i++) { base.BestErrList. } } catch { MessageBox.

MessageBoxIcon. } catch { MessageBox. success = false.Show("Error running Cascade testing!".OK. } base. } public override bool Test() { bool success = true. try { success = CCNet. return success.Test()._te_output).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000150 00000151 00000152 00000153 00000154 00000155 00000156 00000157 00000158 00000159 00000160 00000161 00000162 00000163 00000164 00000165 00000166 00000167 00000168 00000169 00000170 00000171 00000172 00000173 } return success._Tested = success. MessageBoxButtons. } #endregion } Page 140 of 225 . "Error". Report(CCNet.Error).

System. private int _Crossover. //Network Data private double _InputL. _05025397. System.1.“Algorithms\GeneticAlgorithm. private double _OutputL. private int _Mutation. namespace _05025397.Generic.Forms.Collections.Algorithms { public class GeneticAlgorithm : TrainingAlgorithm { #region data /********************\ |* DATA *| \********************/ //GA Data private int _PopSize. //Population Page 141 of 225 . private int _HiddenL.2.5 .Linq. System. ZedGraph. System.Controller.cs” 00000001 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 00000018 00000019 00000020 00000021 00000022 00000023 00000024 00000025 00000026 00000027 00000028 00000029 using using using using using using using using System.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 9.Windows.Text. System.Collections.

//Init Network params double iterations. //Create new population references Population = new ArrayList(_PopSize). #endregion #region constructor /********************\ |* CONSTRUCTOR *| \********************/ public GeneticAlgorithm(int PopSize. out _InputL. int Reports. ITerr) { _PopSize = PopSize. getdata. i++) { Population. Reports. int Mutation. txtData. RichTextBox txtData. RichTextBox txtTest) : base(TargetError.Add Page 142 of 225 . int Iter_Max. int HiddenL. out _OutputL). _Crossover = Crossover. double TargetError. for (int i = 0. bool ITerr. _HiddenL = HiddenL.GetStructure(out iterations. _Mutation = Mutation. int Crossover. txtTest. Iter_Max. i < _PopSize.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000030 00000031 00000032 00000033 00000034 00000035 00000036 00000037 00000038 00000039 00000040 00000041 00000042 00000043 00000044 00000045 00000046 00000047 00000048 00000049 00000050 00000051 00000052 00000053 00000054 00000055 00000056 00000057 00000058 00000059 ArrayList Population.

} //Sort the functions according to fitness now. i < base. j < _PopSize. (int)_OutputL. j++) { ((Model. out results). _HiddenL. double[][] results.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000060 00000061 00000062 00000063 00000064 00000065 00000066 00000067 00000068 00000069 00000070 00000071 00000072 00000073 00000074 00000075 00000076 00000077 00000078 00000079 00000080 00000081 00000082 00000083 00000084 00000085 00000086 00000087 00000088 00000089 (new Model.getMsQE(inputs.GANetwork( (int)_InputL. success = sort_fitnesses().IterMax. for (int i = 0. getdata.GANetwork) Population[j]). double[][] inputs. Page 143 of 225 .Next())). i++) { //Get the msqe for (int j = 0. } } #endregion #region internal_functions /********************\ |* INTERNAL *| \********************/ private bool run() { bool success = true.GetScaledTrainingData(out inputs. Rand. results).

GANetwork)Population[0]).GANetwork)Population[0]). } return success.Report("Best error at iteration [ " + i + " ] was " + ((Model. //Check for having reached the target if (((Model. } success = build_generation(). ((Model.TargetError) { Report("\nNetwork matching or improving upon target error"+ "found at iteration [ " + i + " ].MSqE <= base. } base. break.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000090 00000091 00000092 00000093 00000094 00000095 00000096 00000097 00000098 00000099 00000100 00000101 00000102 00000103 00000104 00000105 00000106 00000107 00000108 00000109 00000110 00000111 00000112 00000113 00000114 00000115 00000116 00000117 00000118 00000119 //Report if we're at a report iteration //Also update the besterrlist (for graphing) if (base. base.CurIt == ReportInterval) { base.MSqE).BestErrList.CurIt++.Add ((double)i."). base.\n").MSqE + ".CurIt = 0. } private bool sort_fitnesses() { Page 144 of 225 .GANetwork)Population[0]).

William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000120 00000121 00000122 00000123 00000124 00000125 00000126 00000127 00000128 00000129 00000130 00000131 00000132 00000133 00000134 00000135 00000136 00000137 00000138 00000139 00000140 00000141 00000142 00000143 00000144 00000145 00000146 00000147 00000148 00000149 bool success = true. MessageBoxButtons. for (int i = (_PopSize . } return success. success = false. selection2.Error). } private bool build_generation() { bool success = true. int selection1.Show("Error sorting population".Sort(). double[] weightarray_f. i--) { //This algorithm is altered from previous versions. double[] weightarray1. "Error". MessageBoxIcon.OK. //It selects two points from the start of the population //Then applies crossover if necessary (based on probability) Page 145 of 225 . try { Population. } catch { MessageBox. double[] weightarray2.1). i > (int)(_PopSize / 2).

} try { weightarray_f = new double[weightarray1. "Error". MessageBoxIcon.Length) { MessageBox. if (Rand.Length != weightarray2.". selection1 = Rand. the least fit) with this new string.Next(100) < _Crossover) { //Choose a random position in the weight array //to be our crossover point Page 146 of 225 . and //so on.Next((int)_PopSize / 2). all the way to the middle of the population. //Then it replaces the last position in the population //(ie. //Next iteration it replaces the next to last. MessageBoxButtons. selection2 = Rand. //Just a quick error check if (weightarray1.Show("Error: 2D weight array length unequal.GANetwork)Population[selection1]).getWeights().OK.Length].Error).getWeights().GANetwork)Population[selection1]). success = false. weightarray2 = ((Model.Next((int)_PopSize / 2).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000150 00000151 00000152 00000153 00000154 00000155 00000156 00000157 00000158 00000159 00000160 00000161 00000162 00000163 00000164 00000165 00000166 00000167 00000168 00000169 00000170 00000171 00000172 00000173 00000174 00000175 00000176 00000177 00000178 00000179 //applies mutation if necessary (same again) //then if neither mutation or Xover have been applied //chooses position 1 or 2. weightarray1 = ((Model.

//Simple single point crossover of the weights for (int j = 0. } } } if (Rand.0. j++) { if (j < XPos) { weightarray_f[j] = weightarray1[j]. (weightarray1. j++) { weightarray_f[j] += Rand.Length.GANetwork)Population[i]).Length .Next(100) < _Mutation) { //Jiggle the weights in the array //up or down by 0.1)).NextDouble() .5 for (int j = 0.Length. } Page 147 of 225 . } } //Set the weights of the current member of the //population we're on ((Model.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000180 00000181 00000182 00000183 00000184 00000185 00000186 00000187 00000188 00000189 00000190 00000191 00000192 00000193 00000194 00000195 00000196 00000197 00000198 00000199 00000200 00000201 00000202 00000203 00000204 00000205 00000206 00000207 00000208 00000209 int XPos = Rand.5. j < weightarray1. } else { weightarray_f[j] = weightarray2[j]. j < weightarray_f.setWeights(weightarray_f).Next(1.

MessageBoxIcon.Append( string. _Crossover)). _Mutation)). output.OK.Append( string.Error).Format("\t\t\tMutation Probability: {0:d}".Append( Page 148 of 225 . output. } } return success. MessageBoxButtons. success = false.Show("Error building new generation". } #endregion #region external_functions /********************\ |* EXTERNAL *| \********************/ public override string network_details() { //Add stringbuilder for efficiency here StringBuilder output = new StringBuilder(). _PopSize)).Format("\t\t\tCrossover Probability: {0:d}".Append( string.Format("Network Details: \n Population: {0:d}".William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000210 00000211 00000212 00000213 00000214 00000215 00000216 00000217 00000218 00000219 00000220 00000221 00000222 00000223 00000224 00000225 00000226 00000227 00000228 00000229 00000230 00000231 00000232 00000233 00000234 00000235 00000236 00000237 00000238 00000239 catch { MessageBox. "Error". output. output.

_HiddenL)).Format("\t\t\tMaximum Generation: {0:d}".Append( string._Trained = success. base.".ToString().Append( string. success = run(). return success. output.00000}".Format("\t\t\t\t\t\t\tIgnore target error (process all iterations): {0}".TargetError)). output.IterMax)). output.Format("\nHidden layer: {0:d}". base. output. return output. } public override bool Test() Page 149 of 225 .Format("\nReport every {0:d} generations. } public override bool Train() { bool success = true. base. _ITerr)).ReportInterval)).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000240 00000241 00000242 00000243 00000244 00000245 00000246 00000247 00000248 00000249 00000250 00000251 00000252 00000253 00000254 00000255 00000256 00000257 00000258 00000259 00000260 00000261 00000262 00000263 00000264 00000265 00000266 00000267 00000268 00000269 string.Append( string.Format("\t\t\tTarget Error: {0:0. base.Append( string.

out tstoutput). j++) { output. double[][] tstinput. double[][] tstoutput.Run(tstinput[i]).Length. netoutput[j])). j++) { output. tstinput[i][j])). getdata. for (int j = 0.Length.GetScaledTestingData(out tstinput.Append("\n\n\t\t\t~~Network Testing~~").GANetwork)Population[0]).Format(" [{0:g}] ". j < netoutput. netoutput = ((Model.Append(string. j < tstinput[i]. i++) { ((Model.GetOutput(). double[] netoutput.Length. output. StringBuilder output = new StringBuilder().William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000270 00000271 00000272 00000273 00000274 00000275 00000276 00000277 00000278 00000279 00000280 00000281 00000282 00000283 00000284 00000285 00000286 00000287 00000288 00000289 00000290 00000291 00000292 00000293 00000294 00000295 00000296 00000297 00000298 00000299 { bool success = true.GANetwork)Population[0]).Format(" [{0:g}] ". Page 150 of 225 .Append(string. for (int j = 0.Append("\nOutput achieved was:"). try { for (int i = 0. i < tstinput.Append("\nWith inputs:"). output. } output.

Append(string.tstoutput). output.getMsQE(tstinput.GANetwork)Population[0]). success = false. } } output. return success. } catch { MessageBox.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000300 00000301 00000302 00000303 00000304 00000305 00000306 00000307 00000308 00000309 00000310 00000311 00000312 00000313 00000314 00000315 00000316 00000317 00000318 00000319 00000320 00000321 00000322 00000323 00000324 00000325 00000326 00000327 00000328 } } output.Format(" [{0:g}] ". } #endregion } Page 151 of 225 .Append("\nOutput Desired was:"). MessageBoxIcon. "Error". base.Append("\nMean Squared Error with these inputs and outputs is:").Report(output. ((Model. MessageBoxButtons.ToString()).".Format ("{0:g}".GANetwork)Population[0]).Length.MSqE)). tstoutput[i][j])). (double) ((Model. } base. j++) { output.Show("Error Running Test.OK.Append( string._Tested = success. j < tstoutput[i]. for (int j = 0.Error).

Linq. private double _LearnRate. System. System.Collections. _05025397. System.Algorithms { public class BackProp : TrainingAlgorithm { #region data /********************\ |* DATA *| \********************/ private double _InputL. private int _HiddenL. #endregion #region constructor Page 152 of 225 .William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 9.Controller.Windows.“Algorithms\BackProp.1.Generic. System. FANN_BackProp BPnet. System.cs” 00000001 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 00000018 00000019 00000020 00000021 00000022 00000023 00000024 00000025 00000026 00000027 00000028 00000029 using using using using using using using using using using System. namespace _05025397. ZedGraph. private double _OutputL. System.2.6 . FANN_Wrapper.Collections.Forms.Text.IO.

double LearnRate. saveDllData(txtData. double TargetError. base. HiddenL. Iter_Max. (int)_OutputL. base. txtTest). _LearnRate. double iterations. _LearnRate = LearnRate. ITerr) { _HiddenL = HiddenL.IterMax).TargetError. RichTextBox txtTest) : base(TargetError. } #endregion #region internal_functions /********************\ Page 153 of 225 . out _OutputL). getdata. BPnet = new FANN_BackProp((int)_InputL. base.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000030 00000031 00000032 00000033 00000034 00000035 00000036 00000037 00000038 00000039 00000040 00000041 00000042 00000043 00000044 00000045 00000046 00000047 00000048 00000049 00000050 00000051 00000052 00000053 00000054 00000055 00000056 00000057 00000058 00000059 /********************\ |* CONSTRUCTOR *| \********************/ public BackProp (int HiddenL. bool ITerr. RichTextBox txtData. out _InputL. int Iter_Max. //Copy the training and testing data to the dat files //ready for the DLL's to access. txtTest. Reports. txtData.ReportInterval.GetStructure(out iterations. int Reports.

tw = new StreamWriter("dlltedata. tw. } #endregion #region external_functions /********************\ |* EXTERNAL *| \********************/ public override string network_details() { StringBuilder output = new StringBuilder(). tw.Format ("Network Details: \n Input Layer: {0:d}".Text). (int) _OutputL)). tw.Close(). output.Append( string.dat").Write(te.Text).Append( Page 154 of 225 . tw.Close(). (int) _InputL)).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000060 00000061 00000062 00000063 00000064 00000065 00000066 00000067 00000068 00000069 00000070 00000071 00000072 00000073 00000074 00000075 00000076 00000077 00000078 00000079 00000080 00000081 00000082 00000083 00000084 00000085 00000086 00000087 00000088 00000089 |* INTERNAL *| \********************/ private void saveDllData(RichTextBox tr.Append( string.Write(tr. RichTextBox te) { TextWriter tw = new StreamWriter("dlltrdata.Format ("\t\t\tOutput Layer: {0:d}". output.dat"). output.

} public override bool Train() { Page 155 of 225 .Format ("\t\t\tHidden Layer: {0:d}".Format ("\nReport every {0:d} generations. output.Format ("\nLearning Rate: {0:0.ReportInterval)). base. output.00000}".Append( string. return output. output. (int) _HiddenL)). output.Format ("\t\t\t\t\t\tIgnore target error " + "(process all iterations): {0}". _LearnRate)). output.Append( string.00000}".". _ITerr)).Append( string.TargetError)).Append( string.ToString(). base.IterMax)).Format ("\t\tMaximum Epoch: {0:d}".Append( string.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000090 00000091 00000092 00000093 00000094 00000095 00000096 00000097 00000098 00000099 00000100 00000101 00000102 00000103 00000104 00000105 00000106 00000107 00000108 00000109 00000110 00000111 00000112 00000113 00000114 00000115 00000116 00000117 00000118 00000119 string. base.Format ("\t\t\tTarget Error: {0:0.

try { double[] besterrlist = BPnet.Error). Report(BPnet.OK. "Error". MessageBoxButtons._Trained = success. } success = getGraphDataFromDll().Add((double)i.Train(). success = false. besterrlist[i]). i < besterrlist.Length. for (int i = 0. MessageBoxIcon.BestErrList. return success.Show("Error running Backprop training!". base. } public bool getGraphDataFromDll() { bool success = true. i++) { base. try { success = BPnet.besterrorlist.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000120 00000121 00000122 00000123 00000124 00000125 00000126 00000127 00000128 00000129 00000130 00000131 00000132 00000133 00000134 00000135 00000136 00000137 00000138 00000139 00000140 00000141 00000142 00000143 00000144 00000145 00000146 00000147 00000148 00000149 bool success = true._tr_output). } catch { MessageBox. Page 156 of 225 .

_te_output). } #endregion Page 157 of 225 .OK. } return success. MessageBoxIcon.Error). } public override bool Test() { bool success = true. return success.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000150 00000151 00000152 00000153 00000154 00000155 00000156 00000157 00000158 00000159 00000160 00000161 00000162 00000163 00000164 00000165 00000166 00000167 00000168 00000169 00000170 00000171 00000172 00000173 00000174 00000175 00000176 00000177 00000178 00000179 } } catch { MessageBox. MessageBoxIcon. success = false. MessageBoxButtons.Error). Report(BPnet.Show("Error getting graph data from DLL".Test(). success = false.Show("Error running Backprop testing!". } catch { MessageBox. try { success = BPnet. "Error". MessageBoxButtons._Tested = success. } base. "Error".OK.

1 .1.Forms.Windows.1. System.Text. System.“Network. System. namespace _05025397.Generic. protected int _numoutputs. protected int _numinputs. protected double[][][] _weights.3.3 .William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000180 00000181 } } 9. _05025397. System. protected double[][] _outputs.Model 9. Page 158 of 225 .Linq.cs” 00000001 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 00000018 00000019 00000020 00000021 00000022 00000023 00000024 using using using using using using System. protected int _hiddennodes.Model { public class Network { #region data /********************\ |* DATA *| \********************/ //[layer][node][input] protected double[][][] _inputs.Collections.

} } private void init_hiddenlayer() { _inputs[1] = new double[_hiddennodes][]. for (int i = 0. #endregion #region init /********************\ |* INIT *| \********************/ private void init_inputlayer() { _inputs[0] = new double[_numinputs][]. Page 159 of 225 . _weights[1][i] = new double[_numinputs + 1]. i < _hiddennodes. i < _numinputs. for (int i = 0. _outputs[1] = new double[_hiddennodes]. _weights[0] = new double[_numinputs][]. _weights[0][i] = new double[2]. _outputs[0] = new double[_numinputs]. i++) { _inputs[1][i] = new double[_numinputs + 1].William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000025 00000026 00000027 00000028 00000029 00000030 00000031 00000032 00000033 00000034 00000035 00000036 00000037 00000038 00000039 00000040 00000041 00000042 00000043 00000044 00000045 00000046 00000047 00000048 00000049 00000050 00000051 00000052 00000053 00000054 protected Random Rand. i++) { _inputs[0][i] = new double[2]. _weights[1] = new double[_hiddennodes][].

0. i++) { _inputs[0][i][1] = 1. Page 160 of 225 . i < _numinputs.0. _weights[2][i] = new double[_hiddennodes + 1].William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000055 00000056 00000057 00000058 00000059 00000060 00000061 00000062 00000063 00000064 00000065 00000066 00000067 00000068 00000069 00000070 00000071 00000072 00000073 00000074 00000075 00000076 00000077 00000078 00000079 00000080 00000081 00000082 00000083 00000084 } } private void init_outputlayer() { _inputs[2] = new double[_numoutputs][].0. i < _numoutputs. _weights[1][i][_numinputs] = 1. _outputs[2] = new double[_numoutputs]. } } private void init_bias() { //Input bias for (int i = 0. for (int i = 0. _weights[2] = new double[_numoutputs][]. } //Hidden layer bias for (int i = 0. i < _hiddennodes.0. i++) { _inputs[1][i][_numinputs] = 1. i++) { _inputs[2][i] = new double[_hiddennodes + 1]. _weights[0][i][1] = 1.

NextDouble(). j++) { _weights[1][i][j] = Rand.0. i++) { _weights[0][i][0] = Rand. i++) { for (int j = 0.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000085 00000086 00000087 00000088 00000089 00000090 00000091 00000092 00000093 00000094 00000095 00000096 00000097 00000098 00000099 00000100 00000101 00000102 00000103 00000104 00000105 00000106 00000107 00000108 00000109 00000110 00000111 00000112 00000113 00000114 } //Output layer bias for (int i = 0. } } private void init_inputweights() { for (int i = 0. } } private void init_hiddenweights() { for (int i = 0. i < _hiddennodes. i < _numoutputs. } } } private void init_outputweights() Page 161 of 225 .NextDouble(). i++) { _inputs[2][i][_hiddennodes] = 1.0. i < _numinputs. j < _numinputs. _weights[2][i][_hiddennodes] = 1.

j++) { _weights[2][i][j] = Rand. -input))). i < _numoutputs. j < _hiddennodes. i < _numinputs. try { //Calculate results for this layer for (int i = 0. } private bool Runinput() { bool success = true. } } } #endregion #region internal /********************\ |* INTERNAL *| \********************/ private double sigmoid(double input) { return (1 / (1 + Math. i++) { for (int j = 0.Pow(Math. double sum = 0.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000115 00000116 00000117 00000118 00000119 00000120 00000121 00000122 00000123 00000124 00000125 00000126 00000127 00000128 00000129 00000130 00000131 00000132 00000133 00000134 00000135 00000136 00000137 00000138 00000139 00000140 00000141 00000142 00000143 00000144 { for (int i = 0.E. i++) { Page 162 of 225 .0.NextDouble().

Show("Error processing input layer". j++) { _inputs[1][i][j] = _outputs[0][j].OK. j++) { sum += (_inputs[0][i][j] * _weights[0][i][j]). j < 2. double sum = 0.Error). sum = 0. i++) { for (int j = 0.0. try { //Feed forward the results from input layer for (int i = 0. "Error". Page 163 of 225 . } return success.0. } } catch { MessageBox. j < _numinputs. MessageBoxButtons. success = false. } private bool Runhidden() { bool success = true. MessageBoxIcon.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000145 00000146 00000147 00000148 00000149 00000150 00000151 00000152 00000153 00000154 00000155 00000156 00000157 00000158 00000159 00000160 00000161 00000162 00000163 00000164 00000165 00000166 00000167 00000168 00000169 00000170 00000171 00000172 00000173 00000174 for (int j = 0. i < _hiddennodes. } _outputs[0][i] = sigmoid(sum).

Error). j++) { sum += (_inputs[1][i][j] * _weights[1][i][j]). } return success. i++) { for (int j = 0. i < _hiddennodes. MessageBoxIcon. } _outputs[1][i] = sigmoid(sum). sum = 0.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000175 00000176 00000177 00000178 00000179 00000180 00000181 00000182 00000183 00000184 00000185 00000186 00000187 00000188 00000189 00000190 00000191 00000192 00000193 00000194 00000195 00000196 00000197 00000198 00000199 00000200 00000201 00000202 00000203 00000204 } } //Calculate results for this layer for (int i = 0. "Error". } private bool Runoutput() { bool success = true.0.0. try Page 164 of 225 . success = false.Show("Error processing hidden layer".OK. double sum = 0. MessageBoxButtons. j < _numinputs + 1. } } catch { MessageBox.

j++) { sum += (_inputs[2][i][j] * _weights[2][i][j]). MessageBoxIcon. i < _numoutputs. j++) { _inputs[2][i][j] = _outputs[1][j]. i++) { for (int j = 0. j < _hiddennodes + 1. success = false.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000205 00000206 00000207 00000208 00000209 00000210 00000211 00000212 00000213 00000214 00000215 00000216 00000217 00000218 00000219 00000220 00000221 00000222 00000223 00000224 00000225 00000226 00000227 00000228 00000229 00000230 00000231 00000232 00000233 00000234 { //Feed forward the results from hidden layer for (int i = 0. } _outputs[2][i] = sigmoid(sum). } return success.Error). } } catch { MessageBox. i++) { for (int j = 0. sum = 0. } } //Calculate results for this layer for (int i = 0.0. MessageBoxButtons.OK. j < _hiddennodes. i < _numoutputs. "Error".Show("Error processing output layer". } Page 165 of 225 .

Page 166 of 225 . MessageBoxIcon. if(success) success = Runinput().Show("Error: Incorrect number of inputs supplied to NN". i++) //Each input node has only one real input //and a bias . success = false. "Incorrect number of inputs".OK. } if (success) for (int i = 0. i < _numinputs.the number of nodes corresponds //to the number of inputs accepted.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000235 00000236 00000237 00000238 00000239 00000240 00000241 00000242 00000243 00000244 00000245 00000246 00000247 00000248 00000249 00000250 00000251 00000252 00000253 00000254 00000255 00000256 00000257 00000258 00000259 00000260 00000261 00000262 00000263 00000264 #endregion #region external /********************\ |* EXTERNAL *| \********************/ public bool Run(double[] inputs) { bool success = true.Length != _numinputs) { MessageBox. MessageBoxButtons. //The numbers of inputs must match up if (inputs. _inputs[0][i][0] = inputs[i]. if(success) success = Runhidden().Error).

} public double[] GetOutput() { //Return the outputs from the //output layer return _outputs[2]. return success. int hiddennodes. int SetRandom) { //Set random number generator Rand = new Random(SetRandom).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000265 00000266 00000267 00000268 00000269 00000270 00000271 00000272 00000273 00000274 00000275 00000276 00000277 00000278 00000279 00000280 00000281 00000282 00000283 00000284 00000285 00000286 00000287 00000288 00000289 00000290 00000291 00000292 00000293 00000294 if(success) success = Runoutput(). _hiddennodes = hiddennodes. int numoutputs. //We'll always have 3 layers Page 167 of 225 . } #endregion #region constructor /********************\ |* CONSTRUCTOR *| \********************/ public Network(int numinputs. _numoutputs = numoutputs. //Set network structure descriptors _numinputs = numinputs.

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3
00000295 00000296 00000297 00000298 00000299 00000300 00000301 00000302 00000303 00000304 00000305 00000306 00000307 00000308 00000309 00000310 } //Input, Hidden, and Output _inputs = new double[3][][]; _weights = new double[3][][]; _outputs = new double[3][]; init_inputlayer(); init_hiddenlayer(); init_outputlayer(); init_bias(); init_inputweights(); init_hiddenweights(); init_outputweights(); } #endregion }

9.1.3.2 - “GANetwork.cs”
00000001 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 using using using using using using using System; System.Collections.Generic; System.Linq; System.Text; System.Windows.Forms; _05025397; System.Collections;

//This class contains Genetic Algorithm specific functions //in addition to the basic feed foward neural net functionality. namespace _05025397.Model

Page 168 of 225

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3
00000013 { 00000014 00000015 00000016 00000017 00000018 00000019 00000020 00000021 00000022 00000023 00000024 00000025 00000026 00000027 00000028 00000029 00000030 00000031 00000032 00000033 00000034 00000035 00000036 00000037 00000038 00000039 00000040 00000041 00000042

public class GANetwork : Network, IComparable { #region data /********************\ |* DATA *| \********************/ double _MSqE; #endregion #region .NET /********************\ |* .NET *| \********************/ //Interface_Implementation //Return Value Meanings: //-<Zero: x < y //-Zero: x == y //->Zero: x > y public int CompareTo(object a) { GANetwork b; if (a is GANetwork) { b = a as GANetwork; return _MSqE.CompareTo(b.MSqE); }

Page 169 of 225

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3
00000043 00000044 00000045 00000046 00000047 00000048 00000049 00000050 00000051 00000052 00000053 00000054 00000055 00000056 00000057 00000058 00000059 00000060 00000061 00000062 00000063 00000064 00000065 00000066 00000067 00000068 00000069 00000070 00000071 00000072 return 0; } #endregion #region external /********************\ |* EXTERNAL *| \********************/ public void getMsQE(double[][] inputs, double[][] outputs) { double sum = 0.0; int counter = 0; double[] netoutputs = new double[_numoutputs]; for (int i = 0; i < inputs.Length; i++) { base.Run(inputs[i]); netoutputs = base.GetOutput(); for (int j = 0; j < netoutputs.Length; j++) { sum += (outputs[i][j] - netoutputs[j]) * (outputs[i][j] - netoutputs[j]); counter++; } } _MSqE = (sum / counter); }

Page 170 of 225

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3
00000073 00000074 00000075 00000076 00000077 00000078 00000079 00000080 00000081 00000082 00000083 00000084 00000085 00000086 00000087 00000088 00000089 00000090 00000091 00000092 00000093 00000094 00000095 00000096 00000097 00000098 00000099 00000100 00000101 00000102 public double[] getWeights() { ArrayList collater = new ArrayList(); //Collect all the weights in an array list //then spit them out as a 1D array of doubles try { for (int i = 0; i < _weights.Length; i++) { for (int j = 0; j < _weights[i].Length; j++) { for (int k = 0; k < _weights[i][j].Length; k++) { collater.Add(_weights[i][j][k]); } } } } catch { MessageBox.Show("Fatal Error collating weights to 1D array", "Error", MessageBoxButtons.OK, MessageBoxIcon.Error); Application.Exit(); } return (double[])collater.ToArray(typeof(double)); } public void setWeights(double[] weights)

Page 171 of 225

i++) { for (int j = 0. MessageBoxButtons. k++.Length. } } } } catch { MessageBox.OK. MessageBoxIcon. Application. j < _weights[i].Exit().Error). k < _weights[i][j]. i < _weights.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000103 00000104 00000105 00000106 00000107 00000108 00000109 00000110 00000111 00000112 00000113 00000114 00000115 00000116 00000117 00000118 00000119 00000120 00000121 00000122 00000123 00000124 00000125 00000126 00000127 00000128 00000129 00000130 00000131 00000132 { //Take a 1D array of doubles and apply //to the correct positions in our weights //array try { for (int i = 0. j++) { for (int k = 0.Show("Fatal Error adding 1D weight array to 3D weight array".Length. } } /********************\ |* GETTERS/SETTERS *| \********************/ public double MSqE { Page 172 of 225 . wc++) { _weights[i][j][k] = weights[wc]. wc = 0. "Error".Length.

} #endregion } Page 173 of 225 . int Nhnodes. int Noutputs.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000133 00000134 00000135 00000136 00000137 00000138 00000139 00000140 00000141 00000142 00000143 00000144 00000145 00000146 00000147 00000148 00000149 00000150 00000151 } get { return _MSqE. SetRandom) { MSqE = 0. int SetRandom) : base(Ninputs. Noutputs.0. } set { _MSqE = value. } } #endregion #region constructor /********************\ |* CONSTRUCTOR *| \********************/ public GANetwork(int Ninputs. Nhnodes.

h” 00000001 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 00000018 00000019 00000020 00000021 00000022 00000023 00000024 00000025 00000026 #pragma once #include "UnManaged_FANN_BackProp. Page 174 of 225 . int OutputL.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 9.DLL Wrapper for FANN functions 9. double TargetErr. namespace FANN_Wrapper { public ref class FANN_Cascade //Managed wrapper for UnManaged_FANN_CasC //which is a C++ interface to the C programmed //FANN dll. using namespace System::Runtime::InteropServices.“FANN_Wrapper. public: /********************\ |* EXTERNAL *| \********************/ FANN_Cascade(int InputL.h" #include "UnManaged_FANN_CasC.4 .4.1.h" using namespace System.1.1 . double LearnRate. { public: /********************\ |* DATA *| \********************/ UnManaged_FANN_CasC* UMWrapper.

William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000027 00000028 00000029 00000030 00000031 00000032 00000033 00000034 00000035 00000036 00000037 00000038 00000039 00000040 00000041 00000042 00000043 00000044 00000045 00000046 00000047 00000048 00000049 00000050 00000051 00000052 00000053 00000054 00000055 00000056 int ReportInterval. } } property array<double>^ besterrorlist { array<double>^ get() MaxNeurons). public: /********************\ |* GETTERS/SETTERS *| \********************/ property String^ _te_output { String^ get() { return gcnew String((wchar_t *) UMWrapper->get_report_test()). !FANN_Cascade(void). Page 175 of 225 . bool Train(void). } } property String^ _tr_output { String^ get() { return gcnew String((wchar_t *) UMWrapper->get_report_train()). int ~FANN_Cascade(void). bool Test(void).

{ public: /********************\ |* DATA *| \********************/ UnManaged_FANN_Backprop* UMWrapper. } } }. i < size. public: Page 176 of 225 . i++) { arout[i] = ar[i]. const double *ar = UMWrapper->get_besterrs(size).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000057 00000058 00000059 00000060 00000061 00000062 00000063 00000064 00000065 00000066 00000067 00000068 00000069 00000070 00000071 00000072 00000073 00000074 00000075 00000076 00000077 00000078 00000079 00000080 00000081 00000082 00000083 00000084 00000085 00000086 { int size. array<double>^ arout = gcnew array<double>(size). for(int i = 0. } return arout. public ref class FANN_BackProp //Managed wrapper for UnManaged_FANN_Backprop //which is a C++ interface to the C programmed //FANN dll.

public: /********************\ |* GETTERS/SETTERS *| \********************/ property String^ _te_output { String^ get() { return gcnew String((wchar_t *) UMWrapper->get_report_test()). double TargetErr. !FANN_BackProp(void). int ReportInterval. bool Train(void). ~FANN_BackProp(void). bool Test(void). } } property String^ _tr_output { String^ get() { return gcnew String((wchar_t *) UMWrapper->get_report_train()). } Page 177 of 225 . int OutputL. int MaximumIteration). double LearnRate. int HiddenL.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000087 00000088 00000089 00000090 00000091 00000092 00000093 00000094 00000095 00000096 00000097 00000098 00000099 00000100 00000101 00000102 00000103 00000104 00000105 00000106 00000107 00000108 00000109 00000110 00000111 00000112 00000113 00000114 00000115 00000116 /********************\ |* EXTERNAL *| \********************/ FANN_BackProp(int InputL.

Page 178 of 225 . const double *ar = UMWrapper->get_besterrs(size).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000117 00000118 00000119 00000120 00000121 00000122 00000123 00000124 00000125 00000126 00000127 00000128 00000129 00000130 00000131 00000132 00000133 00000134 00000135 00000136 00000137 00000138 00000139 } } property array<double>^ besterrorlist { array<double>^ get() { int size. i++) { arout[i] = ar[i]. } return arout. array<double>^ arout = gcnew array<double>(size). for(int i = 0. i < size. } } }.

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3

9.1.4.2 - “FANN_Wrapper.cpp”
00000001 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 00000018 00000019 00000020 00000021 00000022 00000023 00000024 00000025 00000026 00000027 00000028 00000029 #include "stdafx.h" #include "FANN_Wrapper.h" namespace FANN_Wrapper { /********************\ |* FANN_BackProp *| \********************/ //Constructor FANN_BackProp::FANN_BackProp(int InputL, int HiddenL, int OutputL, double LearnRate, double TargetErr, int ReportInterval, int MaximumIteration) { UMWrapper = new UnManaged_FANN_Backprop( InputL, HiddenL, OutputL, LearnRate, TargetErr, ReportInterval, MaximumIteration); } //Destructor FANN_BackProp::~FANN_BackProp(void) { if (UMWrapper) { delete UMWrapper; UMWrapper = 0; } }

Page 179 of 225

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3
00000030 00000031 00000032 00000033 00000034 00000035 00000036 00000037 00000038 00000039 00000040 00000041 00000042 00000043 00000044 00000045 00000046 00000047 00000048 00000049 00000050 00000051 00000052 00000053 00000054 00000055 00000056 00000057 00000058 00000059

//Finalizer FANN_BackProp::!FANN_BackProp(void) { if (UMWrapper) { delete UMWrapper; UMWrapper = 0; } }

//Train bool FANN_BackProp::Train(void) { return UMWrapper->Train(); } //Test bool FANN_BackProp::Test(void) { return UMWrapper->Test(); } /********************\ |* FANN_Cascade *| \********************/ //Constructor FANN_Cascade::FANN_Cascade(int InputL, int OutputL,

Page 180 of 225

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3
00000060 00000061 00000062 00000063 00000064 00000065 00000066 00000067 00000068 00000069 00000070 00000071 00000072 00000073 00000074 00000075 00000076 00000077 00000078 00000079 00000080 00000081 00000082 00000083 00000084 00000085 00000086 00000087 00000088 00000089 double LearnRate, double TargetErr, int ReportInterval, int MaxNeurons) { UMWrapper = new UnManaged_FANN_CasC( InputL, OutputL, LearnRate, TargetErr, ReportInterval, MaxNeurons); } //Destructor FANN_Cascade::~FANN_Cascade(void) { if (UMWrapper) { delete UMWrapper; UMWrapper = 0; } } //Finalizer FANN_Cascade::!FANN_Cascade(void) { if (UMWrapper) { delete UMWrapper; UMWrapper = 0; } } //Train

Page 181 of 225

William Sayers

05025397

Genetic Algorithms and Neural Networks
Milestone 3
00000090 00000091 00000092 00000093 00000094 00000095 00000096 00000097 00000098 00000099 00000100 } bool FANN_Cascade::Train(void) { return UMWrapper->Train(); } //Test bool FANN_Cascade::Test(void) { return UMWrapper->Test(); }

Page 182 of 225

1.“UnManaged_FANN_BackProp.h> #include <fann_cpp.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 9. double _Momentum.4. int _HiddenL.h> class UnManaged_FANN_Backprop //An interface to the C programmed //FANN dll. int _MaximumIteration. int _OutputL.3 . double _TargetErr. //Output for C# Page 183 of 225 . { private: /********************\ |* DATA *| \********************/ //Network Structure int _InputL.h” 00000001 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 00000018 00000019 00000020 00000021 00000022 00000023 00000024 00000025 00000026 00000027 00000028 00000029 #pragma once #include <iostream> #include <string> #include <sstream> #include <vector> #include <doublefann. int _ReportInterval. //Training Parameters double _LearnRate.

00000052 00000053 //Interface functions 00000054 //accessed from C# 00000055 bool Train(void). 00000048 int ReportInterval. int MaximumIteration). 00000047 double LearnRate. int HiddenL. 00000040 00000041 public: 00000042 /********************\ 00000043 |* EXTERNAL *| 00000044 \********************/ 00000045 //Constructor 00000046 UnManaged_FANN_Backprop(int InputL. 00000056 bool Test(void). 00000039 struct fann_train_data *data. int OutputL. 00000032 wchar_t *wchoutput. 00000049 00000050 //Destructor 00000051 ~UnManaged_FANN_Backprop(void). 00000033 00000034 std::vector<double> vecbsterr.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000030 std::wstringstream tr_output. double TargetErr. 00000035 double *bsterr. 00000031 std::wstringstream te_output. 00000057 00000058 public: 00000059 /********************\ Page 184 of 225 . 00000036 00000037 //FANN Data 00000038 struct fann *ann.

const double* get_besterrs(int &size). Page 185 of 225 .William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000060 00000061 00000062 00000063 00000064 00000065 00000066 }. const wchar_t* get_report_test(void). |* GETTERS/SETTERS *| \********************/ const wchar_t* get_report_train(void).

} //Destructor UnManaged_FANN_Backprop::~UnManaged_FANN_Backprop() { fann_destroy_train(data). _TargetErr = TargetErr.h" //Constructor UnManaged_FANN_Backprop::UnManaged_FANN_Backprop (int InputL. int HiddenL. Page 186 of 225 .1. _OutputL = OutputL.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 9. _HiddenL = HiddenL. bsterr = NULL.“UnManaged_FANN_BackProp.h" #include "UnManaged_FANN_BackProp. _MaximumIteration = MaximumIteration. _LearnRate = LearnRate. data = NULL. int MaximumIteration) { _InputL = InputL. wchoutput = NULL.cpp” 00000001 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 00000018 00000019 00000020 00000021 00000022 00000023 00000024 00000025 00000026 00000027 00000028 00000029 #include "StdAfx. int ReportInterval.4 . _ReportInterval = ReportInterval. ann = NULL.4. double TargetErr. int OutputL. double LearnRate.

bsterr = NULL. _OutputL). bool success = true.dat"). double error.) { throw(""). wchoutput = NULL. int reportcounter = 0. } //Train bool UnManaged_FANN_Backprop::Train(void) { static int firstrun = false. ann = fann_create_standard(3. success = false..William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000030 00000031 00000032 00000033 00000034 00000035 00000036 00000037 00000038 00000039 00000040 00000041 00000042 00000043 00000044 00000045 00000046 00000047 00000048 00000049 00000050 00000051 00000052 00000053 00000054 00000055 00000056 00000057 00000058 00000059 delete bsterr. } catch(.. _HiddenL. } if (fann_num_input_train_data(data) != _InputL) Page 187 of 225 . _InputL. delete wchoutput. try { data = fann_read_train_from_file("dlltrdata.

1). //Same range the GA's weights are //initialised too Page 188 of 225 . success = false. FANN_SIGMOID). //Standard backprop fann_set_training_algorithm(ann. fann_set_activation_function_hidden (ann. fann_set_activation_function_output (ann. FANN_TRAIN_BATCH). //Set the learning rate fann_set_learning_rate(ann. success = false. 1).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000060 00000061 00000062 00000063 00000064 00000065 00000066 00000067 00000068 00000069 00000070 00000071 00000072 00000073 00000074 00000075 00000076 00000077 00000078 00000079 00000080 00000081 00000082 00000083 00000084 00000085 00000086 00000087 00000088 00000089 { throw(""). } fann_set_activation_steepness_hidden(ann. fann_set_activation_steepness_output(ann. //Sigmoid Activation Functions (the same one //the GA uses). FANN_SIGMOID). } if (fann_num_output_train_data(data) != _OutputL) { throw(""). (float) _LearnRate).

push_back(error). i++) { //Train one epoch then check the //mean squared error error = (double) fann_train_epoch(ann.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000090 00000091 00000092 00000093 00000094 00000095 00000096 00000097 00000098 00000099 00000100 00000101 00000102 00000103 00000104 00000105 00000106 00000107 00000108 00000109 00000110 00000111 00000112 00000113 00000114 00000115 00000116 00000117 00000118 00000119 fann_randomize_weights(ann.". i < _MaximumIteration. 1.\n".0). if (error < _TargetErr) { tr_output << "\nNetwork matching or improving upon target error" << "found at iteration [ " << i << " ].0. } reportcounter++. if (reportcounter == _ReportInterval) { tr_output << "Best error at iteration [ " << i << " ] was " << error << ". 0. //Training Loop for (int i = 0. vecbsterr. i = _MaximumIteration + 1. reportcounter = 0. data). Page 189 of 225 . } } fann_destroy_train(data).

fann_scale_train_data(data. j++) { te_output << " [" << data->input[i][j] << "] ". } //Test bool UnManaged_FANN_Backprop::Test(void) { fann_type *calc_out = NULL. for (int j = 0. j++) { Page 190 of 225 . return true. 0. i++) { calc_out = fann_run(ann. j < _InputL. i < fann_length_train_data(data). for (unsigned int i = 0. te_output << "\n\n\t\t\t~~Network Testing~~". for (int j = 0. te_output << "\nWith inputs". 1).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000120 00000121 00000122 00000123 00000124 00000125 00000126 00000127 00000128 00000129 00000130 00000131 00000132 00000133 00000134 00000135 00000136 00000137 00000138 00000139 00000140 00000141 00000142 00000143 00000144 00000145 00000146 00000147 00000148 00000149 firstrun = true. error = new double[fann_length_train_data(data)]. j < _OutputL.dat"). } te_output << "\nOutput achieved was". data = fann_read_train_from_file("dlltedata. double *error = NULL. data->input[i]).

} te_output << "\nOutput Desired was". delete calc_out.length() + 1].str(). j++) { te_output << " [" << data->output[i][j] << "] ". wcscpy(wchoutput. Page 191 of 225 . j < _OutputL. } wchoutput = new wchar_t[tr_output. return true.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000150 00000151 00000152 00000153 00000154 00000155 00000156 00000157 00000158 00000159 00000160 00000161 00000162 00000163 00000164 00000165 00000166 00000167 00000168 00000169 00000170 00000171 00000172 00000173 00000174 00000175 00000176 00000177 00000178 00000179 te_output << " [" << calc_out[j] << "] ". for (int j = 0. tr_output. data). wchoutput = NULL.c_str()). te_output << fann_test_data(ann.str(). } } te_output << "\nMean Squared Error with these inputs and outputs is:". } //get_report const wchar_t* UnManaged_FANN_Backprop::get_report_train(void) { if (wchoutput != NULL) { delete wchoutput. delete error.

str().size()]. return wchoutput.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000180 00000181 00000182 00000183 00000184 00000185 00000186 00000187 00000188 00000189 00000190 00000191 00000192 00000193 00000194 00000195 00000196 00000197 00000198 00000199 00000200 00000201 00000202 00000203 00000204 00000205 00000206 00000207 00000208 00000209 return wchoutput. te_output.size(). } const wchar_t* UnManaged_FANN_Backprop::get_report_test(void) { if (wchoutput != NULL) { delete wchoutput. } wchoutput = new wchar_t[te_output. wcscpy(wchoutput.length() + 1]. } //get_besterrs const double* UnManaged_FANN_Backprop::get_besterrs(int &size) { if (bsterr == NULL) { delete bsterr. Page 192 of 225 . bsterr = NULL.c_str()).str(). wchoutput = NULL. size = vecbsterr. } bsterr = new double[vecbsterr.

//Training Parameters Page 193 of 225 .5 .h” 00000001 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 00000018 #pragma once #include <iostream> #include <string> #include <sstream> #include <vector> #include <doublefann.h> class UnManaged_FANN_CasC { public: /********************\ |* DATA *| \********************/ //Network Structure int _InputL. return bsterr. } vecbsterr.1. i < vecbsterr.size(). 9. i++) { bsterr[i] = vecbsterr[i].4.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000210 00000211 00000212 00000213 00000214 00000215 00000216 00000217 00000218 00000219 } for (unsigned int i = 0. int _OutputL.“UnManaged_FANN_CasC.clear().

00000026 std::wstringstream te_output. 00000034 struct fann_train_data *data.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000019 double _LearnRate. Page 194 of 225 . 00000048 int ReportInterval. 00000020 double _TargetErr. 00000022 int _MaxNeurons. int MaxNeurons). 00000023 00000024 //Output for C# 00000025 static std::wstringstream tr_output. double TargetErr. 00000047 double LearnRate. int OutputL. 00000031 00000032 //FANN Data 00000033 struct fann *ann. 00000021 int _ReportInterval. 00000043 00000044 //Constructor 00000045 UnManaged_FANN_CasC 00000046 (int InputL. 00000030 double *bsterr. 00000035 00000036 public: 00000037 /********************\ 00000038 |* EXTERNAL *| 00000039 \********************/ 00000040 00000041 //Destructor 00000042 ~UnManaged_FANN_CasC(void). 00000027 wchar_t *wchoutput. 00000028 00000029 static std::vector<double> vecbsterr.

00000059 00000060 public: 00000061 /********************\ 00000062 |* GETTERS/SETTERS *| 00000063 \********************/ 00000064 const wchar_t* get_report_train(void). 00000057 unsigned int max_epochs. 00000065 const wchar_t* get_report_test(void). unsigned int epochs_between_reports. 00000066 00000067 const double* get_besterrs(int &size). struct fann_train_data *train. 00000053 bool Test(void). 00000058 float desired_error.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000049 00000050 //Interface functions 00000051 //accessed from C# 00000052 bool Train(void). Page 195 of 225 . unsigned int epochs). 00000054 00000055 static int FANN_API Report_Callback 00000056 (struct fann *ann. 00000068 }.

int OutputL. double TargetErr.h" //Static variable declarations std::wstringstream UnManaged_FANN_CasC::tr_output. int ReportInterval. } //Destructor UnManaged_FANN_CasC::~UnManaged_FANN_CasC(void) Page 196 of 225 . _ReportInterval = ReportInterval.“UnManaged_FANN_CasC. ann = NULL.cpp” 00000001 00000002 00000003 00000004 00000005 00000006 00000007 00000008 00000009 00000010 00000011 00000012 00000013 00000014 00000015 00000016 00000017 00000018 00000019 00000020 00000021 00000022 00000023 00000024 00000025 00000026 00000027 00000028 00000029 #include "StdAfx. int MaxNeurons) { _InputL = InputL.6 . std::vector<double> UnManaged_FANN_CasC::vecbsterr.h" #include "UnManaged_FANN_CasC.4. _MaxNeurons = MaxNeurons. double LearnRate. bsterr = NULL. data = NULL. _LearnRate = LearnRate. _OutputL = OutputL.1. wchoutput = NULL.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 9. //Constructor UnManaged_FANN_CasC::UnManaged_FANN_CasC (int InputL. _TargetErr = TargetErr.

wcscpy(wchoutput. } //get_report const wchar_t* UnManaged_FANN_CasC::get_report_train(void) { if (wchoutput != NULL) { delete wchoutput. } const wchar_t* UnManaged_FANN_CasC::get_report_test(void) { if (wchoutput != NULL) { delete wchoutput.length() + 1].str().str(). Page 197 of 225 .William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000030 00000031 00000032 00000033 00000034 00000035 00000036 00000037 00000038 00000039 00000040 00000041 00000042 00000043 00000044 00000045 00000046 00000047 00000048 00000049 00000050 00000051 00000052 00000053 00000054 00000055 00000056 00000057 00000058 00000059 { delete bsterr. wchoutput = NULL. return wchoutput. bsterr = NULL. } wchoutput = new wchar_t[tr_output. wchoutput = NULL.c_str()). fann_destroy_train(data). delete wchoutput. tr_output.

return bsterr. } Page 198 of 225 .str().length() + 1].size()]. size = vecbsterr.clear(). i++) { bsterr[i] = vecbsterr[i].c_str()). } //get_besterrs const double* UnManaged_FANN_CasC::get_besterrs(int &size) { if (bsterr == NULL) { delete bsterr.size(). for (unsigned int i = 0. bsterr = NULL.size(). } wchoutput = new wchar_t[te_output. wcscpy(wchoutput. return wchoutput. te_output.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000060 00000061 00000062 00000063 00000064 00000065 00000066 00000067 00000068 00000069 00000070 00000071 00000072 00000073 00000074 00000075 00000076 00000077 00000078 00000079 00000080 00000081 00000082 00000083 00000084 00000085 00000086 00000087 00000088 00000089 wchoutput = NULL. } vecbsterr.str(). i < vecbsterr. } bsterr = new double[vecbsterr.

. Page 199 of 225 . 00000100 } 00000101 catch(. (float)_LearnRate). 2. 00000118 00000119 fann_set_quickprop_mu(ann. _OutputL). 00000113 00000114 //Some more network customisation here 00000115 //might be nice in future.dat")..) 00000102 { 00000103 throw(""). 00000116 00000117 fann_set_quickprop_decay(ann. 00000096 00000097 try 00000098 { 00000099 data = fann_read_train_from_file("dlltrdata.0). 0. _InputL.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000090 //Train 00000091 bool UnManaged_FANN_CasC::Train() 00000092 { 00000093 int reportcounter = 0. 00000111 00000112 fann_set_learning_rate(ann. 00000104 } 00000105 00000106 if (fann_num_input_train_data(data) != _InputL) 00000107 throw(""). 00000108 00000109 if (fann_num_output_train_data(data) != _OutputL) 00000110 throw("").0). 00000094 00000095 ann = fann_create_shortcut(2.

00000135 00000136 fann_randomize_weights(ann.01f). 1. 1). FANN_SIGMOID). 00000144 00000145 return true.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000120 00000121 fann_set_cascade_weight_multiplier(ann.0). _ReportInterval. 1). (float) _TargetErr). 00000131 00000132 fann_set_activation_function_hidden(ann.0. 0. 00000146 } 00000147 00000148 //Callback for reporting 00000149 int FANN_API UnManaged_FANN_CasC::Report_Callback Page 200 of 225 . 00000133 00000134 fann_set_activation_function_output(ann. 00000139 fann_set_train_stop_function(ann. 0. UnManaged_FANN_CasC::Report_Callback). 00000142 00000143 fann_cascadetrain_on_data(ann. _MaxNeurons. data. FANN_SIGMOID). 150). 00000129 00000130 fann_set_training_algorithm(ann. FANN_STOPFUNC_MSE). 00000124 00000125 fann_set_activation_steepness_output(ann. (float) _LearnRate). 1). 00000137 fann_set_cascade_output_change_fraction(ann. 00000122 00000123 fann_set_cascade_max_out_epochs(ann. 00000138 fann_set_cascade_candidate_change_fraction(ann. 0. FANN_TRAIN_QUICKPROP).01f). 00000140 00000141 fann_set_callback(ann. 00000126 fann_set_activation_steepness_hidden(ann. 00000127 00000128 fann_set_learning_rate(ann.

". vecbsterr.\n". } //Test bool UnManaged_FANN_CasC::Test() { fann_type *calc_out = NULL. Page 201 of 225 . fann_scale_train_data(data. if (MSqE <= desired_error) { tr_output << "\nNetwork matching or improving upon target error" << "found at node [ " << node << " ]. return 0. } node++. unsigned int max_epochs. unsigned int epochs_between_reports.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000150 00000151 00000152 00000153 00000154 00000155 00000156 00000157 00000158 00000159 00000160 00000161 00000162 00000163 00000164 00000165 00000166 00000167 00000168 00000169 00000170 00000171 00000172 00000173 00000174 00000175 00000176 00000177 00000178 00000179 (struct fann *ann. static int node = 1.dat"). float desired_error. double *error = NULL. struct fann_train_data *train. 0. data = fann_read_train_from_file("dlltedata. 1). tr_output << "Best error at node [ " << node << " ] was " << MSqE << ".push_back(MSqE). unsigned int epochs) { double MSqE = fann_get_MSE(ann).

} te_output << "\nOutput Desired was:". i < fann_length_train_data(data). j < _OutputL. i++) { calc_out = fann_run(ann. } te_output << "\nOutput achieved was:". } } te_output << "\nMean Squared Error with these inputs and outputs is:". Page 202 of 225 .William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000180 00000181 00000182 00000183 00000184 00000185 00000186 00000187 00000188 00000189 00000190 00000191 00000192 00000193 00000194 00000195 00000196 00000197 00000198 00000199 00000200 00000201 00000202 00000203 00000204 00000205 00000206 00000207 00000208 00000209 fann_set_callback(ann. te_output << fann_test_data(ann. j++) { te_output << " [" << data->output[i][j] << "] ". data->input[i]). j++) { te_output << " [" << calc_out[j] << "] ". j++) { te_output << " [" << data->input[i][j] << "] ". j < _OutputL. NULL). for (int j = 0. for (int j = 0. for (int j = 0. for (unsigned int i = 0. te_output << "\n\n\t\t\t~~Network Testing~~". error = new double[fann_length_train_data(data)]. te_output << "\nWith inputs:". j < _InputL. data).

delete calc_out. Page 203 of 225 . return true.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 00000210 00000211 00000212 00000213 00000214 00000215 } delete error.

UML Class Diagram The class diagram on the following pages represents the structure of the various classes within the program and how they are combined together to create a whole.2 .William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 9. Page 204 of 225 .

William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Page 205 of 225 .

William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Page 206 of 225 .

William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 -BackProp Previous Page -CascadeCorrelation 1 FANN_Cascade -UMWrapper +FANN_Cascade(in AllDataToFillAttributes) +Train() : bool +Test() : bool 1 -UMWrapper 1 FANN_BackProp +FANN_BackProp(in AllDataToFillAttributes) +Train() : bool +Test() : bool 1 1 1 UnManaged_FANN_BackProp -_InputL : int -_HiddenL : int -_OutputL : int -_LearnRate : double -_TargetErr : double -_ReportInterval : int -_MaximumIteration : int -_Momentum : double -tr_output -te_output -wchoutput -vecbsterr -bsterr -ann -data +UnManaged_FANN_Backprop(in AddDataToFillAttributes) +~UnManaged_FANN_Backprop() +Train() : bool +Test() : bool UnManaged_FANN_CasC -_InputL : int -_OutputL : int -_LearnRate : double -_TargetErr : double -_ReportInterval : int -_MaxNeurons : int -tr_output -te_output -wchoutput -vecbsterr -bsterr -ann -data +UnManaged_FANN_CasC(in AllDataToAFillAttributes) +~UnManaged_FANN_CasC() +Train() : bool +Test() : bool +Report_Callback(in Attributes) : int 1 1 1 1 FANN_Library 9.mit.edu/ga/ o Accessed: 31/10/2008 Page 207 of 225 .Libraries Researched • GALib o Matthew Wall o http://lancet.3 .

sourceforge.net/ o Accessed: 31/10/2008 • FANN (Fast Artificial Neural Network Library) o Steffen Nissen Page 208 of 225 . Franklin o http://ieee.html o Accessed: 31/10/2008 • Flood o Roberto Lopez o http://www.edu/au/~daniel/software/libneural/ o Accessed: 31/10/2008 • Libann o Anonymous o http://www.cimne.nongnu.uow.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 • Libneural o D.org/libann/index.com/flood/ o Accessed: 31/10/2008 • Annie o Asim Shankar o http://annie.

uk/pub/PRNN/README.edu/stat/data/Fisher.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 o http://leenissen.php o Accessed: 31/10/2008 9.ac.math.ox.Datasets • XOR 1 0 1 0 1 1 1 1 0 0 0 0 • • Fishers Iris Data o http://www.4 .stats.uah.dk/fann/index.xhtml Virus Classification o http://www.html Page 209 of 225 .

Research current information on training neural networks using genetic algorithms.5 . and in which cases these methods are suitable. Using above application. Build an application that is capable of training a neural network with a genetic algorithm and with a backpropagation system. and that allows manual adjustment of relevant variables. Deliverable three: A document detailing and explaining the evaluation of different training techniques.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 9. evaluate training methods for neural networks. Page 210 of 225 . Identify the most effective ways of training neural networks using genetic algorithms. as well as previous research. and the conclusions derived from that evaluation. Build a user-interface for the aforementioned application allowing for easy alteration of appropriate variables. Deliverable two: An application meeting the requirements of objectives three and four.Objectives FINAL YEAR DEGREE PROJECT Project Objectives & Deliverables • • • • • • Research the fields of genetic algorithms and neural networks. • • • Deliverable one: A research document describing and detailing the research done for objectives one and two. and a document describing its implementation and usage.

from School of Computer Science. Retrieved April 13. from http://www.cmu.ps Lysle.Works Cited Barber. (2007.com/UploadFile/scottlysle/PassData01142007013005AM/PassData. Bourg. 2009. Retrieved April 13. Bourg. Carnegie Mellon: http://www2. S. AI: Neural Network for Beginners pt. E. Retrieved April 16.cmu.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 10 .cs. Neural Network in C++. 1.pdf Fahlman. Curved trajectory prediction using a self-organizing neural network.gz Darwin.de/~jbr/Papers/GaNNOverview. S. from http://www.aspx Blum. The Cascade-Correlation Learning Architecture. from School of Computer Science. AI for Game Developers (pp.cs. Retrieved April 16. (1859). 2009. M.aifb. 2009. D. G. Wiley Professional Computing. & G. Carnegie Mellon: http://www-2. Graphics processing unit computation of neural networks.csharpcorner. (2006). 2009. or the preservation of favoured races in the struggle for life.edu/user/sef/www/publications/qp-tr.edu/~chris2d/papers/CED_thesis. Passing Data between Windows Forms.unm. (1991).unikarlsruhe.com/KB/recipes/NeuralNetwork_1. (1992). from C# Corner: http://www. M. J.edu/afs/cs. January 16). (1995).ps.cs. Retrieved April 16. O'Reilly.edu/afs/cs. (2001). C.cmu. Branke. S. E.edu/user/sef/www/publications/cascor-tr. C. Retrieved April 13. (1988). 2009. Seeman..codeproject. E.. London: John Murray. S. On the origin of the species by means of natural selection. Chapter 14: Neural Networks. & Lebiere.ps Fahlman. In D. 356-358). A. (2004). & Seeman. 2009.aspx Page 211 of 225 . C.cmu. Davis. from Codeproject: www. An Empirical Study of Learning Speed in Back-Propagation Networks.

Retrieved April 15. from MSDN: http://msdn.org/content/2002/bp. Invention Report S81-64. Retrieved April 16.microsoft.com/robotcontrol. Curved trajectory prediction using a selforganizing neural network. (2009). An introduction to genetic algorithms. Microsoft. (2008).). A. Page 212 of 225 .componentmodel. W. (2002).generation5. (1996).asp McCulloch. D.gz Matthews. 5. 2009. 2009.com/en-us/library/2x8kf7zx. & Srikanth. Genetic algorithms and evolutionary computation. (2004). W.htm Rosenblatt.. Stanford University.UNC-Chapel Hill: http://www. A logical calculus of the ideas immanent in nervous activity. 2009. F. from FANN: http://leenissen.cs.microsoft. Learning Logic. Psychological Review . from http://www. Bulletin of Mathematical Biophysics . from Artificial Neural Networks: http://www.com/en-us/library/system. Office of Technology Licensing. (1958). from generation5: http://www. Nissen. 2009.backgroundworker. Retrieved April 15. Back-propagation for the Uninitiated. 2009. (2009). Massachusetts: MIT Press. B.talkorigins. The perceptron: A probabilistic theory for information storage and organization in the brain. & Pitts.unc. V. Retrieved 21 01.ps. J. Robot Control.d.aspx Microsoft..html Marshall.learnartificialneuralnetworks. Using C++ Interop (Implicit PInvoke). (n. 386-408. BackgroundWorker Class.aspx Mitchell. 65. from MSDN: http://msdn. M. (1982). 2009. Retrieved April 13. J.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Marczyk. 115-133.dk/fann/ Parker. A. from Computer Science . (1999). (1943).com/faqs/genalg/genalg. S.edu/~marshall/WWW/PAPERS/curve9912. 2009. Retrieved April 15. Retrieved April 16.

ic. C. from http://zedgraph. (1974).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Stergiou. from Imperial College London: http://www. Retrieved April 15.org/wiki/index. PhD.ac. & Dimitrios. P. ZedGraph. (n.uk/~nd/surprise_96/journal/vol4/cs11/report.html#Contents Werbos. 2009. (1997). Retrieved 21 01. S. Beyond Regression: New tools for prediction and analysis in the behavioral sciences..php?title=Main_Page Page 213 of 225 .). 2009. Neural Networks.doc.d. Harvard University.

from Codeproject: http://www. Retrieved April 13. S. 1. P. (2006). August 6). & G. Davis.ps. Bourg. 2009.. from http://www. AI: Neural Network for Beginners pt.cs. 2009. M.pdf DiLascia. S.php/c9855/ Page 214 of 225 .com/KB/recipes/Backprop_ANN. Neural Network in C++. G. Retrieved April 13.aspx Blum. Wiley Professional Computing. from Codeguru: http://www. E.codeproject. J. (2006). AI: Neural Network for Beginners pt. (2004). from MSDN: http://msdn. (2007. Retrieved April 13.aifb. or the preservation of favoured races in the struggle for life. C.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 11 . 2009. 2.codeproject.com/Cpp/Cpp/cpp_mfc/tutorials/article.edu/~chris2d/papers/CED_thesis. On the origin of the species by means of natural selection.codeproject. 2009. 2009. Retrieved April 13. Curved trajectory prediction using a self-organizing neural network. from Codeproject: www. Retrieved April 13.com/enus/magazine/cc301501. In D. AI: Neural Network for Beginners pt. (1992). 356-358). Chapter 14: Neural Networks. 3. Retrieved April 13.unm. (2001). from Codeproject: http://www. AI for Game Developers (pp. Bourg. London: John Murray.com/KB/cs/GA_ANN_XOR. O'Reilly. (2002). D.unikarlsruhe. from http://www. & Seeman. (1995). Killing Processes Cleanly. (2006).de/~jbr/Papers/GaNNOverview.com/KB/recipes/NeuralNetwork_1. 2009.codeguru. (1859).aspx Barber. Retrieved April 13. S. Seeman. Call Unmanaged DLLs from C#. Branke.aspx DLL Tutorial for Beginners.gz Darwin.microsoft. C.aspx Barber. 2009. M.Bibliography Barber. A. Graphics processing unit computation of neural networks.

cs.cmu. January 16). J.talkorigins. from School of Computer Science.edu/afs/cs. Back-Propagation: CBPNet.ps. from School of Computer Science. Retrieved April 13. from XORGA: http://www. Retrieved April 16. Carnegie Mellon: http://www2.csharpcorner. from http://code. Retrieved 21 01.aspx Marczyk. Retrieved April 15.edu/~marshall/WWW/PAPERS/curve9912. Retrieved April 15.generation5. Retrieved April 16.com/p/googlechartsharp/ Lysle. The Cascade-Correlation Learning Architecture. (1988). 2009. An Empirical Study of Learning Speed in Back-Propagation Networks. generation5. (2004).asp McCulloch.UNC-Chapel Hill: http://www. A logical calculus of the ideas immanent in nervous activity. 2009.generation5.google. from generation5: http://www. Curved trajectory prediction using a selforganizing neural network. from generation5: http://www.cmu. 2009. Page 215 of 225 .William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Fahlman. (n.cs. 115-133..org/content/2000/cbpnet. E.ps GoogleChartSharp. (2002). (2007.gz Matthews. (2000). & Pitts.html Marshall. Passing Data between Windows Forms. & Srikanth. S. (1991). J.org/content/2002/bp. 2009.d. A. Retrieved April 15. J.asp Matthews. from C# Corner: http://www.cmu. Back-propagation for the Uninitiated.com/UploadFile/scottlysle/PassData01142007013005AM/PassData. A..cs. from http://www.. Carnegie Mellon: http://www-2. 2009. (1999).cmu. S. 2009.unc.edu/user/sef/www/publications/qp-tr. Retrieved 26 01. E. S.). 2009. W. from Computer Science . C.edu/afs/cs. Bulletin of Mathematical Biophysics .com/faqs/genalg/genalg.d. 2009. & Lebiere. (1943). (n. J.generation5.ps Fahlman. 5.). Genetic algorithms and evolutionary computation. 2009. Retrieved April 16.asp Matthews. W. V.edu/user/sef/www/publications/cascor-tr.org/content/2000/xorga.

from Datatypes: http://leenissen.microsoft.).org/nplot/wiki/ Parker. (n.d.backgroundworker. M. B. Nguyen.dk/fann/html/files/fann_data-h. Retrieved 21 01.. S.d. Retrieved 21 01. Neural Networks. (n. (2009). (n. An introduction to genetic algorithms. 2009. (1990). Nissen. S. Invention Report S81-64. (n. 2009.).d.d.dk/fann/ Nissen. from FANN: http://leenissen. from MSDN: http://msdn. 2009.html Nplot Charting Library for . 1990. & Widrow.com/en-us/library/2x8kf7zx.html#fann_train_enum Nissen. Improving the learning speed of 2-layer neural networks by choosinginitial values of the adaptive weights.d. (1996). BackgroundWorker Class.html Nissen. 2009.aspx Mitchell. Robot Control.html Nissen. Using C++ Interop (Implicit PInvoke). from FANN: http://leenissen.com/en-us/library/system. 2009. from http://netcontrols. Stanford University. B. Reference Manual. 1990 IJCNN International Joint Conference on . Massachusetts: MIT Press.).learnartificialneuralnetworks. Office of Technology Licensing. Learning Logic. from MSDN: http://msdn.). 21-26.).William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Microsoft.htm Page 216 of 225 .componentmodel.). 2009. (2008).aspx Microsoft. 2009. Retrieved 21 01. struct fann_train_data.dk/fann/fann_1_2_0/r1597. S. from FANN: http://leenissen. (n. Retrieved 21 01. S. D. (n.microsoft.d. (2009). 2009. (1982). Retrieved April 16.. from Artificial Neural Networks: http://www.com/robotcontrol. from FANN: http://leenissen.NET. Retrieved April 16. Retrieved 26 01.dk/fann/fann_1_2_0/r1837. struct fann. Retrieved 21 01.dk/fann/html/files/fann-h. S. Retrieved April 15. D. 2009.

.................. Retrieved 21 01.. The Cascade-Correlation Learning Architecture.... 10 Figure 2 ....... 1991)................... 13 Figure 4 ... F.....)....................... The perceptron: A probabilistic theory for information storage and organization in the brain........ from http://zedgraph..... P....ac........ PhD............ ........ (1997)..................... C.....local optimum (units are undefined in this case). Stergiou................ 27 Figure 5 ............ S............. 2009.......................... ZedGraph. Retrieved April 15......ic................. Psychological Review .html#Contents Werbos......A human neuron ...The sigmoid function shape ...... Harvard University..Tables 12.. 65. 2009...d..................................................William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Rosenblatt..doc......................... (1974).................. Neural Networks... & Dimitrios.............Table of Figures Figure 1 ................ 386-408.... The circular connections are adjustable weights (Fahlman & Lebiere.....1 ..... .. (n...... (1958)......... 35 Page 217 of 225 ..................... Beyond Regression: New tools for prediction and analysis in the behavioral sciences..org/wiki/index............. from Imperial College London: http://www..An artificial neuron ....... 12 Figure 3 ....................php?title=Main_Page 12 .......The initial state of a cascade correlation neural network......A feed-forward neural network ....uk/~nd/surprise_96/journal/vol4/cs11/report... 30 Figure 6 ..................

..............................................................William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Figure 7 ....Learning Algorithm tab ....... 48 Figure 21 ..............Dataset display design (About form) ....Mean squared error graph tab .................. 35 Figure 8 .. The CascadeCorrelation Learning Architecture....................... 51 Page 218 of 225 ......................Dataset tab design (Main form) .............. 42 Figure 12 ...............The second state of a cascade correlation neural network............ 47 Figure 20 ...........Design of the training algorithm tab (Wizard form) ..................... 41 Figure 10 ................................................... with two nodes added (Fahlman & Lebiere..................................................... 49 Figure 22 .......The third state of a cascade correlation neural network................ 50 Figure 23 ............................................. 36 Figure 9 – Network and output tab design (Main form) ...... The Cascade-Correlation Learning Architecture.............. The square connections are locked weights(Fahlman & Lebiere...........................Network Settings tab...........About form design (About form) .................Network and Output tab................Licenses display design (About form) ............................................................................................................................................Dataset selection tab design (Wizard form)..................................... 42 Figure 11 ...................... 45 Figure 17 ......................... 1991)................... ... 43 Figure 13 .....................................Design of the network settings tab (Wizard form) ....... A first node has been added................................ 1991)....................The dataset tab ............................................................................... 46 Figure 18 .................. 45 Figure 16 .Working form design (Working form) .........................Mean squared error tab design (Main form) ............ 44 Figure 14 ................................................... 44 Figure 15 ................... 46 Figure 19 ...............

...................... 53 Figure 27 .............................Number of iterations to achieve target (XOR) .......Mean squared error at testing (Viruses).. 64 Figure 34.............................. 61 Figure 31 .. 65 Figure 35 ............................................. 54 Figure 29 .....................Working form ................................... 68 12..........Mean squared error upon last training test (Viruses)................Mean squared error upon last training test (XOR) ...A sigmoid function(Bourg & Seeman............................. 52 Figure 26 .........The mean squared error at testing (XOR).Summation of the inputs and weights ..... 12 Equation 2 ................About form................................ Licenses view ..........................................About form....................... ............... Back-propagation for the Uninitiated......................... 51 Figure 25 ....................... main view . as defined in the graph above..............Number of iterations to achieve target (Viruses) ..... 61 Figure 30 .....................William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Figure 24 ............................. 62 Figure 32 .................The mean squared error at testing (Iris data) ............2 ......... 54 Figure 28 .......................................... 67 Figure 37 ...Dataset tab ......................... In this example “y” is the function output and “t” is the function input...... 67 Figure 36 .................... 64 Figure 33........................Number of iterations to achieve target (Iris data) . 13 Equation 3 – The delta rule(Matthews...Table of Equations Equation 1 .....................................Mean squared error upon last training test (Iris data) ..................................................................................................... 2002) .................................... 21 Page 219 of 225 ................................ 2004)..............................About form. Datasets view ..........

......................................... ‘f'p’ is the derivative for pattern 'p' of the candidate unit's activation function with respect to the sum of it’s inputs and ‘Ii......................................................................................differentiated sigmoid function (Matthews............... Back-propagation for the Uninitiated.......Altered delta calculation rule for the hidden layer(s) (Matthews.............. 34 12...............The definition of 's' given that 'o' is the network output at which the error is measured........................ 22 Equation 8 ..................... 'p' is the training pattern................................................................................The statistical approach to deciding relevant data .................................. ............. 'v' is the candidate unit's value......... The CascadeCorrelation Learning Architecture.......... 59 Table 3 ...................Altered delta calculation rule for the output layer(Matthews................ Backpropagation for the Uninitiated............. 59 Page 220 of 225 ........Table of tables Table 1 .... 34 Equation 10 ...............Back-propagation testing data (XOR)................................The partial derivative of ‘s’ with respect to each of the candidates incoming weights ‘wi’......................... Backpropagation for the Uninitiated.......... 1991)...............The partial derivative of ’s’ (Fahlman & Lebiere................................................................. 2002) ........ Back-propagation for the Uninitiated...................Altered perceptron learning rule (Matthews.............................William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Equation 4 ........................................... 21 Equation 7 ................. 2002)..............................3 . 2002) ......p’ is the input the candidate receives from unit ‘i’ for pattern ‘p’.................. where ‘oo’ is the sign of the correlation between the candidates value output ‘o’.... 33 Equation 9 .. 15 Table 2 ..................... ....... The Cascade-Correlation Learning Architecture.... the quantities ‘`v ’ and ‘`Eo’ are the values of v and Eo averaged over all patterns (Fahlman & Lebiere................................. 21 Equation 6 .....Genetic Algorithm testing data (XOR) .... 21 Equation 5 ........................................................ 1991)... 2002)..

...............................Genetic algorithm testing data (Iris data) .. 63 Table 7 ...............Cascade Correlation testing data (Viruses) ......................... 66 Table 9 ....................................Cascade correlation testing data (Iris data) ...... 63 Table 6 ......................Cascade Correlation testing data (XOR) .....Back-propagation testing data (Iris data) ...Back-propagation testing data (Viruses) ...................................................................................................... 63 Table 8 .Genetic Algorithms testing data (Viruses) ............................... 60 Table 5 ....................................William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 Table 4 ............................... 66 Page 221 of 225 ................................ 66 Table 10 ..............

o Decided on XOR as a starting point. o Submitted objectives. • 24/10/2008 o University computers were down so code demo was impossible. o Decision made to adjust the weights of the neural network during training as opposed to using a genetic algorithm to establish network structure. 17/10/2008 o Investigating more deeply into journals based on research already conducted.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 13 . • 10/10/2008 o GALib decided as a genetic algorithm starting point o Further non-web references as a goal • • 16/10/2008 o Investigation into various neural network libraries taking place. • 31/10/2008 o First presentation complete and ready o References building up o Project plan complete • 10/11/2008 o Presentation went well o Progress satisfactory • 14/11/2008 o Discussed document formatting and layout Page 222 of 225 . o Agreed on further investigation into GALib and genetic algorithms.Diary • 01/10/2008 – Research progress satisfactory. however. back-propagation network was running and learning XOR.

21/11/2008 o Progress satisfactory 28/11/2008 o Progress satisfactory 05/12/2008 o Progress satisfactory 12/12/2008 o Progress satisfactory 09/01/2009 o Progress satisfactory 16/01/2009 o Progress satisfactory 23/01/2009 o Progress satisfactory 30/01/2009 o Milestone two has been successfully submitted o Started investigating other areas to explore outside of project requirements.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 o Progress satisfactory. o Progress satisfactory • 20/02/2009 o Progress satisfactory Page 223 of 225 . • • 06/02/2009 o Still exploring extra areas for project to cover. • • • • • • • • • 19/11/2008 o Milestone one submitted successfully. 13/02/2009 o Have decided to extend the project by including cascade correlation learning algorithm and comparison to genetic algorithms.

have decided to implement in managed language (C#) to try to solve this o Report continues satisfactorily • 04/04/2009 o To do: About box Add dataset notes Decided on virus classification as a third dataset via the custom dataset functionality. is much faster (fixed several bugs along the way) and retains none of the problematic memory leaks. o Report satisfactory • 05/04/2009 o Further work on user interface needed o Finish C# genetic algorithm code (C++ code does work but is very slow) o Report satisfactory • 06/04/2009 o C# genetic algorithm code is complete.William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 • • • • 27/02/2009 o Progress satisfactory 06/03/2009 o Project progress delayed by other coursework 13/03/2009 o Project progress delayed by other coursework 20/03/2009 o Project progressing. catch-up time for previous two weeks will be used over Easter break. Page 224 of 225 . o Have started report • 02/04/2009 o Main problems experienced with genetic algorithm seem related to memory leaks.

Page 225 of 225 .William Sayers 05025397 Genetic Algorithms and Neural Networks Milestone 3 o Report satisfactory • 10/04/2009 o Fully functional project – commencing report. o Report satisfactory.

Sign up to vote on this title
UsefulNot useful