%----------------------------------------------------------------------------
%
% Hopfield network
% 
% function [x,W] = hopfield(inp,pats,num_iter,ref)
%
%	inp: A column vector containing the input
%
%	pats: Matrix containing patterns to be memorized.
%	      Each row corresponds to one memory pattern.
%
%	num_iter: number of iterations to simulate the Hopfield dynamics
%
%	ref: index value of the target pattern (row number in the matrix pats)
%
%
%	x : returned settled pattern
%
%	W : returned weight
%
% example:
%
% 1. Noisy input
%
% g = []; for i=1:10; g =[g; vec(dogfilter(rand(20,20),17))'>0]; end; inp=g(5,:).*(rand(1,20*20)<0.6); imagesc(reshape(inp,20,20)); inp=inp*2-1; g=g*2-1; [X,w] = hopfield(inp',g,100,5);
%
% g = []; for i=1:10; g =[g; vec(dogfilter(rand(20,20),17))'>0]; end; inp=(g(5,:)*2-1).*((rand(1,20*20)<0.2)*2-1); imagesc(reshape(inp,20,20)); g=-(g*2-1); [X,w] = hopfield(inp',g,100,5);
%
% 2. Partial input
%
% g = []; for i=1:10; g =[g; vec(dogfilter(rand(20,20),17))'>0]; end; inp=reshape(g(5,:),20,20); inp(1:10,:)=zeros(10,20); inp=vec(inp)'; imagesc(reshape(inp,20,20)); inp=inp*2-1; g=g*2-1; [X,w] = hopfield(inp',g,100,5);
%
% Author: Yoonsuck Choe
% http://faculty.cs.tamu.edu/choe
% License: GNU Public License http://www.gnu.org
%
%----------------------------------------------------------------------------
function [x,W] = hopfield(inp,pats,num_iter,ref)

%----------------------------------------
% 1. Book-keeping
%----------------------------------------

[N,n] = size(pats);

m = sqrt(n);

%----------------------------------------
% 2. Determine weights (learning)
%----------------------------------------

W = zeros(n,n);

for i=1:N
  % Fill this in: You need pats.
  W = ..............
end

% Remove self-feedback

W = W .* (1-eye(n,n));

%----------------------------------------
% 3. Activate 
%----------------------------------------

% x will be the evolving state vector, initialized to the input pattern
x = inp;

% x0 is a snapshot at time 0 (this will not change)
x0 = inp;

% Main loop
for i=1:num_iter

   for k=1:n
	
	%--------------------
	% 3.1 pick a unit at random, for updating
	%--------------------
   	pick = ceil(rand*n);

	%--------------------
	% 3.2 determine the new state of the neuron just picked
	%	Fill this in: you need W, pick, and x.
	%	Note: the new value should be either -1 or 1.
	%--------------------
   	x(pick) = ......................

	%--------------------
	% 3.3 Plot results: you may need to tinker with this a bit.
	%	
	%--------------------
        loc=find(x>=0); locx = floor(loc/m); locy = rem(loc,m);
        loc0=find(x0>=0); locx0 = floor(loc0/m); locy0 = rem(loc0,m);
        rloc=find(pats(ref,:)>=0); rlocx = floor(rloc/m); rlocy = rem(rloc,m);
    
        if (i==1 && k==1) 
            figure(1);
	    %-- uncomment line below for octave
            % gset("xrange [-1:41]"); gset("pointsize 2");
            plot(locx0,m-locy0,'gx',rlocx+m+1,m-rlocy,'rx');
        else 
            figure(2);
	    %-- uncomment line below for octave
            % gset("xrange [-1:41]"); gset("pointsize 2");
            plot(locx,m-locy,'bx');
        end

   end

end