Bayes’ theorem as a classifier

Bayes’ theorem allows us to compute the combined a posteriori estimate of the probability that the measurement indicates that the object is in any one of multiple classes. We can then select the class with maximum a posteriori probability. Those computations, however, require a large amount of information on covariance matrices that cannot always be provided. Here is a way to estimate those covariance matrices and thus enable at least an approximate Bayesian analysis.

http://www.ece.purdue.edu/~landgreb/SaldjuCovarEst.pdf