Professional Documents
Culture Documents
Soft Computing PDF
Soft Computing PDF
1 2
3 4
1
Guiding Principles of Soft
Premises of Soft Computing
Computing
• The real world problems are pervasively • The guiding principle of soft computing is:
imprecise and uncertain – Exploit the tolerance for imprecision,
• Precision and certainty carry a cost uncertainty, partial truth, and approximation to
achieve tractability, robustness and low
solution cost.
5 6
7 8
2
Current Applications using Soft
Unique Property of Soft computing
Computing
• Learning from experimental data • Application of soft computing to handwriting recognition
• Application of soft computing to automotive systems and
• Soft computing techniques derive their manufacturing
power of generalization from • Application of soft computing to image processing and
data compression
approximating or interpolating to produce • Application of soft computing to architecture
outputs from previously unseen inputs by • Application of soft computing to decision-support
using outputs from previous learned inputs systems
• Application of soft computing to power systems
• Generalization is usually done in a high- • Neurofuzzy systems
dimensional space. • Fuzzy logic control
9 10
11 12
3
Neural Networks Overview Definitions of Neural Networks
• Neural Network Definition • According to the DARPA Neural Network Study
(1988, AFCEA International Press, p. 60):
• Some Examples of Neural Network
• ... a neural network is a system composed of
Algorithms and Architectures
many simple processing elements operating in
• Successful Applications parallel whose function is determined by network
structure, connection strengths, and the
processing performed at computing elements or
nodes.
13 14
15 16
4
Definitions of Neural Networks A Simple Neural Network
• According to Zurada (1992), p. xv:
• Artificial neural systems, or neural
networks, are physical cellular systems
which can acquire, store, and utilize
experiential knowledge.
The network has 2 inputs, and one output. All are binary.
The output is
1 if W0 *I0 + W1 * I1 + Wb > 0
0 if W0 *I0 + W1 * I1 + Wb <= 0
We want it to learn simple OR: output a 1 if either I0 or I1 is 1.
17 18
19 20
5
A Simple Neural Network A Simple Perceptron Network
• We train the network on these examples. Weights after The decision surface is a line in this case.
each epoch (exposure to complete set of patterns) This restricts the possible mappings
• At this point (8) the network has finished learning. Since available.
(D-Y)=0 for all patterns, the weights cease adapting.
Single perceptrons are limited in what they can learn:
• If we have two inputs, the decision surface is a line and
its equation is I1 = (W0/W1).I0 + (Wb/W1)
• In general, they implement a simple hyperplane decision
surface
• This restricts the possible mappings available.
21 22
6
Multilayer Perceptron Multilayer Perceptron
• The hidden layer learns to recode (or to • The units are a little more
complex than those in the
provide a representation for) the inputs. original perceptron: their
More than one hidden layer can be used. input/output graph is as
shown in the left.
• The architecture is more powerful than • Output Y as a function:
single-layer networks: it can be shown that Y = 1 / (1+ exp(-k.(Σ Win
* Xin))
any mapping can be learned, given two • The graph shows the
hidden layers (of units). output for k=0.5, 1, and
10, as the activation
varies from -10 to 10.
25 26
7
Radial Basis function Networks Training RBF Networks
• Radial basis function networks • Generally, the hidden unit
are also feedforward, but have function is a Gaussian:
only one hidden layer. • Gaussians with three different
• Like MLP, RBF nets can learn standard deviations as shown
arbitrary mappings: the primary in the left figure.
difference is in the hidden Training RBF Networks.
layer. • RBF networks are trained by
• RBF hidden layer units have a • deciding on how many hidden
receptive field which has a units there should be
centre: that is, a particular
input value at which they have • deciding on their centers and
a maximal output. Their output the sharpness (standard
tails off as the input moves deviation) of their Gaussians
away from this point.
29 30
31 32
8
Architectures for Processing Time-
Unsupervised networks
series
• Simple Perceptrons, • In an unsupervised net, the network adapts purely in
response to its inputs. Such networks can learn to pick
MLP, and RBF out structure in their input.
networks need a Applications for unsupervised nets
teacher to tell the • clustering data:
– exactly one of a small number of output units comes on in
network what the response to an input.
desired output should • reducing the dimensionality of data:
– data with high dimension (a large number of input units) is
be. These are compressed into a lower dimension (small number of output
supervised networks. units).
• Although learning in these nets can be slow, running the
trained net is very fast - even on a computer simulation
of a neural net.
33 34
35 36
9
Support Vector Machines and
Neural Network Applications
Neural Networks
• Pen PC's
The learning machine that uses data to find the
– PC's where one can write on a tablet, and the APPROXIMATING FUNCTION (in regression
writing will be recognised and translated into problems) or the SEPARATION BOUNDARY (in
(ASCII) text. classification, pattern recognition problems), is
the same in high-dimensional situations.
• Speech and Vision recognition systems It is either the so-called SVM or the NN.
– Not new, but Neural Networks are becoming
increasingly part of such systems. They are
used as a system component, in conjunction
with traditional computers.
37 38
39 40
10
SVMs and NNs Introduction of SVM
41 42
43 44
11
Optimal Hyperplane Optimal Hyperplane (Cont’d)
45 46
47 48
12
Optimal Hyperplane (cont’d) Optimal Hyperplane (cont’d)
49 50
51 52
13
Characteristics of Solutiions Geometrical Interpretation
53 54
55 56
14
Nonlinear Separation SVM: A Hyperplane Classifier
57 58
59 60
15
Some SVM Features
SVMs combine three important ideas
• Applying optimization algorithms from
Operations Research (Linear
Programming and Quadratic
Programming)
• Implicit feature transformation using
kernels
• Control of overfiting by maximizing the
margin
61
16