TrainKernel.m

function [Y2hat,results,model] = TrainKernel(K,K2,Y,Y2,D,p,gamma,e,C,regression)

if strcmpi(regression,'krr')

    H = K + gamma*eye(size(K));
    model = H\Y(p+1:end-D);
    Y2hat = K2*model; 
    svs = 100;
    
elseif strcmpi(regression,'svr')

  	% -----------------------------------------------------------------------------	
	% Training
	% -----------------------------------------------------------------------------
	H = K;	% Standard SVR
	model = svmtrain(Y(p+1:end-D),H,['-t 4 -s 3 -c ' num2str(C) ' -p ' num2str(e)]);
	svs  = 100*length(model.SVs)/length(Y(p+1:end-D));
    
	% -----------------------------------------------------------------------------	
	% Testing
	% -----------------------------------------------------------------------------
	H2 = K2;	% Standard SVR
	Y2hat = svmpredict(Y2(p+1:end-D),H2,model);

elseif strcmpi(regression,'rvm')
	
    % Training
	H = K; YT  = Y(p+1:end-D);
	% Testing
	H2 = K2; 
    
    % Fixed hyperparameters
	alpha 	= 1; 
	beta 	= 1e3; 
	maxIts  = 1000;
	monIts  = 1;
	[weights, used] = sbl_estimate(H,YT,alpha,beta,maxIts,monIts);
	
	PHI2     = H2(:,used);
	Y2hat    = PHI2*weights;
	svs      = 100*length(weights~=0)/length(YT);	
    model.weights = weights; model.used = used; model.H=H;model.H2=H2; model.YT=YT;
end;
	
% --------------------------------------------------------------------------------------------------------------------------
% Results in test
% --------------------------------------------------------------------------------------------------------------------------
results = ComputeResults(Y2hat,Y2(p+1:end-D));
results.svs = svs;