patterns.m

It can be used to matching the patterns

Ni=10; % Inhibitory neurons as output neurons
Ne=15; % 3by5 pattern i.e., 15 Excitatory neurons as input neurons
a1=0.02; a2=0.06; %1 for exc. and 2 for inh.
b1=0.2; b2=0.22; 
c1=-55; c2=-65;
d1=4; d2=2;
w=[];%initalize the weight from input neuron to output neuron
I=[];%Initialize the current Input for all neurons
firings=[]; % spike timings initialization


%%%%%%%%%%%% Training starts%%%%%%%%%%%%%%%

Input(:,1)=[0 30 0 0 30 0 0 30 0 0 30 0 0 30 0]; % First Input Pattern--1 for exc. inputs
Input(:,2)=[30 30 30 0 0 30 30 30 30 30 0 0 30 30 30]; % Second Input Pattern--2 for exc. inputs
Input(:,3)=[30 30 30 0 0 30 0 30 30 0 0 30 30 30 30]; % Third Input Pattern--3 for exc. inputs
Input(:,4)=[0 0 30 0 30 30 30 0 30 30 30 30 0 0 30]; % Forth Input Pattern--4 for exc. inputs
Input(:,5)=[30 30 30 30 0 0 30 30 30 0 0 30 30 30 30]; % Fifth Input Pattern--5 for exc. inputs
Input(:,6)=[0 30 30 30 0 0 30 30 30 30 0 30 30 30 30]; % Sixth Input Pattern--6 for exc. inputs
Input(:,7)=[30 30 30 0 0 30 0 30 0 30 0 0 30 0 0]; % seventh Input Pattern--7 for exc. inputs
Input(:,8)=[30 30 30 30 0 30 0 30 0 30 0 30 30 30 30]; % Eighth Input Pattern--8 for exc. inputs
Input(:,9)=[30 30 30 30 0 30 30 30 30 0 0 30 30 30 0]; % Eighth Input Pattern--9 for exc. inputs
Input(:,10)=[30 30 30 30 0 30 30 0 30 30 0 30 30 30 30]; % Eighth Input Pattern--0 for exc. inputs

%it is assumed that, for simplicity, for pattern A- 1st neuron fires,
%pattern B- 2nd neruon fires,C-3rd fires and D-4th fires, so you have to
%preselect initial weights for it.

for i=1:15
    for j=1:10
        if Input(i,j)==0
        w(i,j)=-10;
        else
        w(i,j)=4.5;    
        end
    end 
end
w=w'; % Because if one input neuron fires, we can use one col. of w to add with 4 o/p neurons

firings=[]; % spike timings initialization

for in=1:10 %for 8 patterns
    initial=1+(in-1)*10;
    v=-65*ones(Ne+Ni,1); % Initial values of v for each input patterns
    u=[];% variable delcaration, otherwise it will mix with other u of other programs
    u(1:15,1)=b1*v(1:15); % Initial values of u for exc. neurons for each input
    u(16:25,1)=b2*v(16:25,1);% Initial values of u for inh. neurons for each input
    
    for t=initial:initial-1+10 % simulation of 10 ms with one input
        I(16:25,1)=zeros(10,1);%At every time-step you have to initialize it to zero
        I(1:15,1)=Input(1:15,in); %Specify the input for the exc. neuron
        v(1:15,1)=v(1:15,1)+0.5*(0.04*v(1:15,1).^2+5*v(1:15,1)+140-u(1:15,1)+I(1:15,1)); 
        % Excitatory neurons are upgraded with extern. inputs patterns
        for i=1:15 % for excitatory input neurons, if fires, it is updated and produce input I for output neruons
            if v(i,1)>30 % one input neuron fires, 
            firings=[firings; t, i];
            v(i,1)=c1;  %it is updated
            u(i,1)=u(i,1)+d1; %it is updated
            I(16:25,1)=I(16:25,1) + w(1:10,i);%weight from the firing neurons add with the all output neurons
            end
        end
t;
        for j=16:25 % for output neruons, if it fires, then you have to upgrade---it, other o/p neurons and weights
            if ( v(j,1)==max(v(16:23,1)) )&& ( v(j,1)>25 )
                firings=[firings; t, j];
                v(j,1)=c2;    %it is updated
                u(j,1)=u(j,1)+d2;  %it is updated
                for k=16:25 % -ve contribution from this firing neuron to other inhibitory output neurons
                    if k~=j
                    I(k,1)=I(k,1)-50;
                    % -ve contribution from this firing neuron to other inhibitory output neurons
                    end
                end
                
                for m=1:10
                     x=find((firings(:,2)==m)&(firings(:,1)>initial));
                     if ~isempty(x)
                        %if v(m,1)>30 % it means both pre and post fired, and pre is first to fire
                        w(j-15,m)= w(j-15,m) + .2; % the weight is strenthened as STDP rule
                     else % it means pre does not fire but post fire
                        w(j-15,m)= w(j-15,m) - 15; % the weight is weakened as STDP rule
                     end
                 end

             end
        end
%         for i=1:3 % for excitatory input neurons, if fires, it is updated and produce input I for output neruons
%             if v(i,1)>30 % input one neuron fires, 
%             v(i,1)=c1;  %it is updated
%             u(i,1)=u(i,1)+d1; %it is updated
%             end
%         end
        % Now you have to upgrade all the output neurons (input neurons are upgraded beginning of the cycle)
       % v(16:19,1)=v(16:19,1)+0.5*(0.04*v(16:19,1).^2+5*v(16:19,1)+140-u(16:19,1)+I(16:19,1));
        v(16:25,1)=v(16:25,1)+0.5*(0.04*v(16:25,1).^2+5*v(16:25,1)+140-u(16:25,1)+I(16:25,1));
        x(1:10,1)=v(16:25,1);
        w;
        u(16:25,1)=u(16:25,1)+a2*(b2*v(16:25,1)-u(16:25,1));
    t;
    w;
    end
end
plot(firings(:,1),firings(:,2),'.');
axis([0 101 0 30.5])


%%%%%%%%%%%% Testing starts%%%%%%%%%%%%%%%
%     firings=[];
%     v=-65*ones(Ne+Ni,1); % Initial values of v for each input patterns
%     u=[];% variable delcaration, otherwise it will mix with other u of other programs
%     u(1:15,1)=b1*v(1:15); % Initial values of u for exc. neurons for each input
%     u(16:23,1)=b2*v(16:23);% Initial values of u for inh. neurons for each input
%     
%     for t=1:10 % simulation of 10 ms with one input
%         I(16:23,1)=zeros(8,1);%At every time-step you have to initialize it to zero
%         I(1:15,1)=Input(1:15,pattern_input); %Specify the input for the exc. neuron
%         v(1:15,1)=v(1:15,1)+0.5*(0.04*v(1:15,1).^2+5*v(1:15,1)+140-u(1:15,1)+I(1:15,1)); 
%         % Excitatory neurons are upgraded with extern. inputs patterns
%         for i=1:15 % for excitatory input neurons, if fires, it is updated and produce input I for output neruons
%             if v(i,1)>30 % one input neuron fires, 
%             firings=[firings; t, i];
%             v(i,1)=c1;  %it is updated
%             u(i,1)=u(i,1)+d1; %it is updated
%             I(16:23,1)=I(16:23,1) + w(1:8,i);%weight from the firing neurons add with the all output neurons
%             end
%         end
% t
%         for j=16:23 % for output neruons, if it fires, then you have to upgrade---it, other o/p neurons and weights
%             if ( v(j,1)==max(v(16:23,1)) )&& ( v(j,1)>30 )
%                 firings=[firings; t, j];
%                 v(j,1)=c2;    %it is updated
%                 u(j,1)=u(j,1)+d2;  %it is updated
%                 for k=16:23 % -ve contribution from this firing neuron to other inhibitory output neurons
%                     if k~=j
%                     I(k,1)=I(k,1)-50;
%                     % -ve contribution from this firing neuron to other inhibitory output neurons
%                     end
%                 end
%                 
% %                 for m=1:8
% %                      x=find((firings(:,2)==m)&(firings(:,1)>initial));
% %                      if ~isempty(x)
% %                         %if v(m,1)>30 % it means both pre and post fired, and pre is first to fire
% %                         w(j-15,m)= w(j-15,m) + .2; % the weight is strenthened as STDP rule
% %                      else % it means pre does not fire but post fire
% %                         w(j-15,m)= w(j-15,m) - 15; % the weight is weakened as STDP rule
% %                      end
% %                  end
% 
%              end
%         end
% %         for i=1:3 % for excitatory input neurons, if fires, it is updated and produce input I for output neruons
% %             if v(i,1)>30 % input one neuron fires, 
% %             v(i,1)=c1;  %it is updated
% %             u(i,1)=u(i,1)+d1; %it is updated
% %             end
% %         end
%         % Now you have to upgrade all the output neurons (input neurons are upgraded beginning of the cycle)
%        % v(16:19,1)=v(16:19,1)+0.5*(0.04*v(16:19,1).^2+5*v(16:19,1)+140-u(16:19,1)+I(16:19,1));
%         v(16:23,1)=v(16:23,1)+0.5*(0.04*v(16:23,1).^2+5*v(16:23,1)+140-u(16:23,1)+I(16:23,1));
%         x(1:8,1)=v(16:23,1)
%         w
%         u(16:23,1)=u(16:23,1)+a2*(b2*v(16:23,1)-u(16:23,1));
%     t;
%     w;
%     end
% figure;
% %plot(firings(1,1),firings(1,2),'.');
% plot(firings(:,1),firings(:,2),'.');
% axis([0 21 0 25.5])