classiferBlockMean.m

%% Trainig Classfier based on Block Mean Features:

function [Wih, Who] = classiferBlockMean(TrainSample,TrainTarget,TrainCont,Alpha,NH1,Epoch,H,W,WinSize,m)

	%% Training Section For Weights Of Block Mean Features
	
	Wih = (rand(((H/WinSize)*(W/WinSize))+1,NH1)-.5)*.25; % Weights Between Inputs And Hidden Layer
	Who = (rand(NH1+1,m)-.5)*.25; % Weights Between Hidden Layer And Out Put
	
	for epoch = 1:Epoch
		for r = 1:TrainCont
			
			nEt= Wih' * TrainSample(:,r);
			Z= 1./(1+exp(-nEt));  %output for the hidden layer 
			Zb=[1;Z]; %adding a bias
			
			nEt2= Who' * Zb ;
			Output= 1./(1+exp(-nEt2));  %Final outputs
			
			
			% Adjust delta values of weights For output layer:
	        delta_o = Output.*(1-Output).*(TrainTarget(:,r)-Output);
	     
			% Propagate the delta backwards into hidden layers:	     
	        ezb= Who * delta_o;
	        delta_hb = Zb.*(1-Zb).*ezb;
	        delta_h=delta_hb(2:end);
			% Add weight changes to original weights:
	       
	        Who = Who + Alpha*Zb*delta_o';       
	        Wih= Wih + Alpha*TrainSample(:,r)*delta_h';
            
		end
    end
end