#4 TRAINING YOUR NEURAL NETWORK- (PERCEPTRON BASED)

In machine learning, the perceptron is an algorithm for supervised classification of an input into one of several possible non-binary outputs. It is a type of linear classifier, i.e. a classification algorithm that makes its predictions based on a linear predictor function combining a set of weights with the feature vector. The algorithm allows for online learning, in that it processes elements in the training set one at a time.

The perceptron algorithm dates back to the late 1950s. Its first implementation, in custom hardware, was one of the first artificial neural networks to be produced.

PROGRAM

% Prajwal's Neural Network :P :P 

clear all

close all

clc;s=0;ch=0;

e=input('enter the enable input for training (either 1 or 0): ');

l=input('for bipolar inputs, press "-1". For binary inputs, press "0": ');

x=input('enter the input matrix (order: 4x2): ');

t=input('enter the target matrix (order: 4x1): ');

w=input('enter the weight matrix (order: 2x1): ');

b=input('enter the bias: ');

lr=input('enter the learning rate: ');

it=input('enter the no.of iterations to be performed: ');

y=[0; 0; 0; 0];

if(e==1)

 for s=1:it

     ch=0;

    for q=1:4

        yin=b+x(q,1)*w(1,1)+x(q,2)*w(2,1);

        if(yin<0)

            y(q,1)=l;

        else

            y(q,1)=1;

        end

        if(y(q,1)~=t(q,1))

            w(1,1)=w(1,1)+lr*t(q,1)*x(q,1);

            w(2,1)=w(2,1)+lr*t(q,1)*x(q,2);

            b=b+lr*t(q,1);

        else

            ch=ch+1;

        end       

    end

    if ch==4

        break;

    end

 end

end

disp(s);

disp(w);

disp(y);

Conclusion: 

• Training your neural network based on the algorithm specified above comes under SUPERVISED LEARNING.
• In perceptron based learning, we should initialize weights to zero. But you can even initialize them with any random values as shown in simulation.
• If weights aren't initialized with 0, it becomes another different learning algorithm

For any queries regarding code or anything else regarding any post in my blog, contact me at pj.kotamraju225@gmail.com. See you people with a new conclusion within no time ;) :D