Some Useful Multivariate Gaussian Information Quantities

Many information quantities for multivariate normal distributions have nice closed forms. Their essential parts usually reduce to logs of minor determinant quotients so they combine nicely. Here's a list of them. All of them are somewhere in Adaptive Wireless Communications by Bliss and Govindasamy.


All covariance matrices are assumed to be full-rank. By convention the determinant of an empty matrix is 1.

Real Identities

Circularly-symmetric complex normal identities are identical except without the \frac{1}{2} factor.