function demoAndOr(flag)
%function demoAndOr(flag)
% flag: 1 - and, 2- or 3 - xor

%rand('state',0)

% Define input patterns
two_inputs = [0 0; 0 1; 1 0; 1 1];

% Define target vectors
and = [0 0 0 1];
or = [0 1 1 1];
xor = [0 1 1 0];

% Define eta value and maximum number of iterations
eta = 0.1;
maxIters = 200;

% Train network to learn each pattern
if nargin == 0, flag = 1; end
if (flag == 1)
    disp('AND:')
    [w,w0] = learn2D(two_inputs,and,eta,maxIters);
elseif flag == 2
    disp('OR:')
    [w,w0] = learn2D(two_inputs,or,eta,maxIters);
elseif flag == 3
    disp('XOR:')
    [w,w0] = learn2D(two_inputs,xor,eta,maxIters);
end

end %function

function [weights,w0] = learn2D(inputs,target,eta,cycles)
%   [w,w0] = learn2D(x,y,eta,cycles)
%   where
%           x is a vector of input vectors
%           y is a vector of corresponding expected responses
%           eta is the fraction step size for perceptron learning rule
%           cycles is the max number of iteration without convergence before giving up
%   and
%           w is a vector of the weights
%           w0 is the threshold value
N = length(target);
weights = rand(1,length(inputs(1,:)));
w0 = rand;
disp('Initial Weights:');
str = sprintf('\tw0 =%g',w0);
disp(str);
for i=1:length(weights),
    str = sprintf('\tw%g = %g',i,weights(i));
    disp(str);
end

counter = 1;
% Plot Initial
figure
x1 = inputs(:,1);
x2 = inputs(:,2);
x = [-2:0.1:2];
hold on
for i = 1:N,
    if (target(i)==1),
        plot(x1(i),x2(i),'.');
    else,
        plot(x1(i),x2(i),'o');
    end
end
hold off

while counter <= cycles,
    % Use Perceptron Learning Rule to set weights
    for i = 1:N,
        output = perceptron(weights,inputs(i,:),w0);
        delta(i) = target(i)-output;
        weights = weights+eta*delta(i).*inputs(i,:);
        w0 = w0+eta*delta(i);
    end
    if (sum(abs(delta)) == 0),
        str = sprintf('\nConvergence after %d iterations.\n',counter-1);
        disp(str);
        counter = cycles;
        % Plot Final values
        hold on;
        y = -((x*weights(1)+w0)/weights(2));
        plot(x,y,'r');
        axis([-1 2 -1 2]);
        hold off;
        title(str);
        break;
    end
    % Plot incremental values
    if counter<10
        hold on;
        y = -((x*weights(1)+w0)/weights(2));
        plot(x,y,':');
        axis([-1 2 -1 2]);
        hold off;
        pause(1)
    end
    counter = counter + 1;
end

if (counter > cycles),
    str = sprintf('\nFAILURE to converge after %d iterations.\n',counter-1);
    disp(str);
    title(str);
end

% Display results
disp('Final Weights:');
str = sprintf('\tw0 =%g',w0);
disp(str);
for i=1:length(weights),
    str = sprintf('\tw%g = %g',i,weights(i));
    disp(str);
end
testResults(weights,inputs,w0);
end % function

function [v] = perceptron(w,u,w0)
% Perceptron that returns an output for a given input vector,u, with
% weights w and threshold value w0.  The length of vectors u and w must be the
% same.
if (length(w) ~= length(u))
    error('length of w and u must match');
end
if ((w*u'+w0) >= 0) v = 1;
else v = 0;
end
end % function

function testResults(w,u,w0)
% Displays inputs and outputs for 2 node network
str=sprintf('\nResults for given weight and inputs:\nx1  |  x2   |   y');
disp(str);
disp('----|-------|----');
for i=1:length(u(:,1)),
    str = sprintf('%g   |   %g   |   %g',u(i,1),u(i,2),perceptron(w,u(i,:),w0));
    disp(str);
end
str = sprintf('\n');
disp(str);
end