clear all; close all;

%input vectors
p1=[0;1;0];
p2=[1;0;0];
p3=[1;1;0];

%initial weights
w21=ones(3,3); 
w12=0.5*w21;  

%1.	Compute the Layer 1 response:
a1=p1  

%2.	Next, compute the input to Layer 2:
w12*a1  

%Since all neurons have the same input, pick the first neuron as the
%winner. (In case of a tie, pick the neuron with the smallest index.)
a2=[1;0;0]  

%3.	Now compute the L2-L1 expectation:
w211=w21*a2  

%4.	Ajust the Layer 1 output to include the L2-L1 expectation:
a1=p1&w211  
a1=[0;1;0];

%5.	Next, the Orienting Subsystem determines the degree of match between
%the expectation and the input pattern:
norm(a1)^2/norm(p1)^2
%  ans = 1	> p=0.4
%Therefore no reset
a0=0;  

%6.	Since a0=0, continue with step 7.

%7.	Resonance has occurred, therefore update row 1 of w12:
w121=2*a1/(2+norm(a1).^2-1)  
w12(1,:)=w121'  
%8.	Update column 1 of w21:
w21(:,1)=a1  

%9.	Remove p1, and return to step 1 with input pattern p2.
a1=p2  

%2.	Compute the input to Layer 2:
w12*a1  
%Since neurons 2 and 3 have the same input, pick the second neuron as the winner:
a2=[0;1;0]  

%3.	Now compute the L2-L1 expectation:
w212=w21*a2  

%4.	Adjust the Layer 1 output to include the L2-L1 expectation:
a1=p2&w212  
a1=[1;0;0]
%5.	Next, the Orienting Subsystem determines the degree of match between
%the expectation and the input pattern:

norm(a1)^2/norm(p2)^2  
%ans = 1  >p=0.4

%Therefore no reset
a0=0;  

%6.	Since a0=0, continue with step 7.
%7.	Resonance has occurred, therefore update row 2 of w12:
w122=2*a1/(2+norm(a1).^2-1)  

w12(2,:)=w122'  

%8. Update column 2 of w21:
w21(:,2)=a1  

%9.Remove p2, and return to step 1 with input pattern p3.

%1.	Compute the new Layer 1 response:
a1=p3  

%2.	Compute the input to Layer 2:
w12*a1  

%Since all neurons have the same input, pick the first neuron as the winner:
a2=[1;0;0]  

%3.	Now compute the L2-L1 expectation:
w211=w21*a2  

%4.	Adjust the Layer 1 output to include the L2-L1 expectation:
a1=p3&w211  
a1=[0;1;0]
%5.	Next, the Orienting Subsystem determines the degree of match between the expectation and the input pattern:

norm(a1)^2/norm(p3)^2  
%ans = 0.5000  > p=0.4

%Therefore no reset
a0=0;

%6.	Since a0=0, continue with step 7.
%7.	Resonance has occurred, therefore update row 1 of w12:
w121=2*a1/(2+norm(a1).^2-1)  
w12(1,:)=w121'  

%8. Update column 1 of w21:
w21(:,1)=a1  

%This completes the training, since if you apply any of the three patterns again they will not change the weights. These patterns have been successfully clustered. This type of result (stable learning) is guaranteed for the ART1 algorithm, since it has been proven to always produce stable clusters.