Project:

Intelligent systems. Task for the laboratory for Perceptron training.

Aim:

Learn to write training (parameter estimation) algorithm for simple linear classifier.

Tasks:

1. Create a program for classification using single Perceptron of two different objects into two classes according to two estimated features. Output estimation should be done acording to the following equation: y = 1, if x1*w1 + x2*w2 + b > 0; y = -1, if x1*w1 + x2*w2 + b <= 0; here w1, w2 and b are the parameters, which should be initialized randomly and updated (estimated) later, using training algorithm; x1 and x2 are the object features, estimated using functions written in Matlab (available in Matlab code template or in Data.txt).
2. Write a training algorithm for created classifier. Use this parameter update formula: w1(n+1) = w1(n) + eta*e(n)*x1(n); here 0 < eta < 1; e(n) = d(n) - y(n); - an error, estimated by comparing the desired output d(n) and the output, currently (for current example) received from the Perceptron. w2(n+1) = w2(n) + eta*e(n)*x2(n); b(n+1) = b(n) + eta*e(n);

Additional task:

Solve this classification task using Naive Bayes Classifier. Some examples and tutorials for additional task:

• http://www.statsoft.com/textbook/naive-bayes-classifier
• https://en.wikipedia.org/wiki/Naive_Bayes_classifier