MMGDX: a maximum-margin training method for neural networks

by

 

17 Sep 2010 (Updated )

Maximum-margin training method applicable to MLP in the context of binary classification.

[acc,estimated]=sim_MMGDX(Nor,W1,W2,b1,b2,X,t)
function [acc,estimated]=sim_MMGDX(Nor,W1,W2,b1,b2,X,t)
X=Nor*X;
[L,Col]=size(X);
error=0;
for k=1:Col
    estimated(k)=W2*logsig(W1*X(:,k)+b1)+b2;
    estimated_aprox=sign(estimated(k));
    if not(estimated_aprox==t(k))
        error=error+1;
    end
end

acc=(Col-error)/Col;

Contact us