Daniel@0: function Xi = lossHinge(W, Psi, M, gradient)
Daniel@0: %
Daniel@0: %   Xi = lossHinge(W, Psi, M, gradient)
Daniel@0: %   
Daniel@0: %   W:          d*d metric
Daniel@0: %   Psi:        d*d feature matrix
Daniel@0: %   M:          the desired margin
Daniel@0: %   gradient:   if 0, returns the loss value
Daniel@0: %               if 1, returns the gradient of the loss WRT W
Daniel@0:     
Daniel@0:     Xi = max(0, M - sum(sum(W .* Psi)));
Daniel@0: 
Daniel@0:     if gradient & Xi > 0
Daniel@0:         Xi = -Psi;
Daniel@0:     end
Daniel@0: end