annotate toolboxes/FullBNT-1.0.7/KPMstats/KLgauss.m @ 0:e9a9cd732c1e tip

first hg version after svn
author wolffd
date Tue, 10 Feb 2015 15:05:51 +0000
parents
children
rev   line source
wolffd@0 1 function kl = KLgauss(P, Q)
wolffd@0 2 %The following computes D(P||Q), the KL divergence between two zero-mean
wolffd@0 3 %Gaussians with covariance P and Q:
wolffd@0 4 % klDiv = -0.5*(log(det(P*inv(Q))) + trace(eye(N)-P*inv(Q)));
wolffd@0 5
wolffd@0 6 R = P*inv(Q);
wolffd@0 7 kl = -0.5*(log(det(R))) + trace(eye(length(P))-R);
wolffd@0 8
wolffd@0 9 %To get MI, just set P=cov(X,Y) and Q=blockdiag(cov(X),cov(Y)).