Daniel@0: function kl = KLgauss(P, Q)
Daniel@0: %The following computes D(P||Q), the KL divergence between two zero-mean
Daniel@0: %Gaussians with covariance P and Q:
Daniel@0: %  klDiv = -0.5*(log(det(P*inv(Q))) + trace(eye(N)-P*inv(Q)));
Daniel@0: 
Daniel@0: R = P*inv(Q);
Daniel@0: kl = -0.5*(log(det(R))) + trace(eye(length(P))-R);
Daniel@0: 
Daniel@0: %To get MI, just set P=cov(X,Y) and Q=blockdiag(cov(X),cov(Y)).