Professional Documents
Culture Documents
Machine Learning: An Artificial Intelligence Methodology: WWW - Ijecs.in
Machine Learning: An Artificial Intelligence Methodology: WWW - Ijecs.in
in
International Journal Of Engineering And Computer Science ISSN:2319-7242
Volume 2 Issue 12, Dec.2013 Page No. 3400-3404
*2
UG, CSE
Dronacharya Group of Institutions,
Greater Noida
M18nikhilkumar@gmail.com
Abstract— The problem of learning and decision making is at the core level of argument in biological as well as artificial aspects. So
scientist introduced Machine Learning as widely used concept in Artificial Intelligence. It is the concept which teaches machines to
detect different patterns and to adapt to new circumstances. Machine Learning can be both experience and explanation based learning.
In the field of robotics machine learning plays a vital role, it helps in taking an optimized decision for the machine which eventually
increases the efficiency of the machine and more organized way of preforming a particular task. Now-a-days the concept of machine
learning is used in many applications and is a core concept for intelligent systems which leads to the introduction innovative technology
and more advance concepts of artificial thinking.
Anish Talwar 1IJECS Volume 2 Issue 12, Dec. 2013, Page No.3400-3404 Page 3401\
problems faced during the learning process are as Supervised learning is an algorithm in which both
follows:- the inputs and outputs can be perceived. Based on
this training data, the algorithm has to generalize
Bias- the tendency to prefer one hypothesis over such that it is able to correctly respond to all possible
another is called a bias. Consider the agents N and P. inputs. This algorithm is expected to produce correct
Saying that a hypothesis is better than N's or P's output for inputs that weren’t encountered during
hypothesis is not something that is obtained from the training. In supervised learning what has to be
data - both N and P accurately predicts all of the data learned is specified for each example. Supervised
given - but is something external to the data. classification occurs when a trainer provides the
Without a bias, an agent will not be able to make any classification for each example. Supervised learning
predictions on unseen examples. The hypotheses of actions occurs when the agent is given immediate
adopted by P and N disagree on all further examples, feedback about the value of each action.
and, if a learning agent cannot choose some In order to solve a give problem using supervised
hypotheses as better, the agent will not be able to learning algorithm one has to follow some certain
resolve this disagreement. To have any inductive steps:-
process make predictions on unseen data, an agent
requires a bias. What constitutes a good bias is an 1) Determine the type of training examples.
empirical question about which biases work best in 2) Gather a training set.
practice; we do not imagine that either P's or N's 3) Determine the input feature representation of
biases work well in practice. learned function.
4) Determine the structure of learning function
Noise-In most real-world situations, the data are & corresponding learning algorithm.
not perfect. Noise exists in the data (some of the 5) Complete the design and run the learning
features have been assigned the wrong value), there algorithm on the gather set of data.
are inadequate features (the features given do not 6) Evaluate the accuracy of the learned function
predict the classification), and often there are also the performance of the learning function should
examples with missing features. One of the be measured and then the performance should be
important properties of a learning algorithm is its again measured on the set which is different from the
ability to handle noisy data in all of its forms. training set.
Anish Talwar 1IJECS Volume 2 Issue 12, Dec. 2013, Page No.3400-3404 Page 3403\
In this algorithm we have to first select the tree induction, naive Bayes, etc) are examples of
number of cluster in advance, they might converge supervised learning techniques.
to a local minimum. K-means can be seen as a Unsupervised learners are not provided with
specialization of the expectation maximization (EM) classifications. In fact, the basic task of unsupervised
algorithm. It is more efficient (lower computational learning is to develop classification labels
complexity) than hierarchical clustering. automatically. Unsupervised algorithms seek out
Algorithm: similarity between pieces of data in order to
1. K-means ((X= {d1, . . . ,dn} ⊆ Rm , k): 2R) determine whether they can be characterized as
2. C: 2R /*µ a set of clusters */ forming a group. These groups are termed clusters,
3. d = Rm x Rm -> R /*distance function*/ and there are whole families of clustering machine
4. µ: 2R -> R /* µ computes the mean of a learning techniques.
cluster */ In unsupervised classification, often known as
5. select C with k initial centers f1,….fk 'cluster analysis' the machine is not told how the
6. while stopping criterion not true do texts are grouped. Its task is to arrive at some
7. for all clusters cj ∈ C do grouping of the data. In a very common of cluster
8. cj ← {di|∀fld(di, fj) ≤d(di, fl)} analysis (K-means), the machine is told in advance
9. done how many clusters it should form -- a potentially
10. for all means fj do difficult and arbitrary decision to make.
11. fj <- µ(cj) It is apparent from this minimal account that the
12. done machine has much less to go on in unsupervised
13. done classification. It has to start somewhere, and its
14. return C algorithms try in iterative ways to reach a stable
configuration that makes sense. The results vary
widely and may be completely off if the first steps
are wrong. On the other hand, cluster analysis has a
much greater potential for surprising you. And it has
considerable corroborative power if its internal
comparisons of low-level linguistic phenomena lead
to groupings that make sense at a higher
interpretative level or that you had suspected but
deliberately withheld from the machine. Thus cluster
analysis is a very promising tool for the exploration
of relationships among many texts.
VI. CONCLUSIONS & FUTURE SCOPE
Anish Talwar 1IJECS Volume 2 Issue 12, Dec. 2013, Page No.3400-3404 Page 3404\
implemented for the experiments, was able to distinguish [2] Zoubin Ghahramani,"Unsupervised Learning",Gatsby Computational
Neuroscience Unit, University College London, UK.
between two classifiers that were similar in terms of accuracy
but different in terms of classifier complexity. Since time is [3] Rich Caruana; Alexandru Niculescu- Mizil,"An Empirical Comparison of
often of essence when evaluating, e.g., if the evaluation method Supervised Learning Algorithms", Department of Computer Science, Cornell
is used as a fitness function for a genetic algorithm, we have University, Ithaca, NY 14853 USA
analyzed measure-based evaluation in terms of the time
[4] Niklas Lavesson,"Evaluation and Analysis of Supervised Learning
consumed to evaluate different classifiers. The conclusion is Algorithms and Classifiers", Blekinge Institute of Technology Licentiate
that the evaluation of lazy learners is slower than for eager Dissertation Series No 2006:04,ISSN 1650-2140,ISBN 91-7295-083-8
learners, as opposed to cross-validation tests. Additionally, we
[5] Yugowati Praharsi; Shaou-Gang Miaou; Hui-Ming Wee,” Supervised
have presented a method for measuring the impact that learning
learning approaches and feature selection - a case study in diabetes”,
algorithm parameter tuning has on classifier performance using International Journal of Data Analysis Techniques and Strategies 2013 - Vol. 5,
quality attributes. The results indicate that parameter tuning is No.3 pp. 323 - 337
often more important than the choice of algorithm. Quantitative
[6] Andrew Ng,"Deep Learning And Unsupervised","Genetic Learning
support is provided to the assertion that some algorithms are
Algorithms","Reinforcement Learning and Control", Department of Computer
more robust than others with respect to parameter configuration. Science,Stanford University,450 Serra Mall, CA 94305, USA.
Anish Talwar 1IJECS Volume 2 Issue 12, Dec. 2013, Page No.3400-3404 Page 3405\