Kybernetika 31 no. 4, 331-335, 1995

Minimum entropy of error estimate for multi-dimensional parameter and finite-state-space observations

Antonín Otáhal


The minimum entropy of error estimate (MEEE) is studied for a finite mixture of probability densities on a finite-dimensional Euclidean space. It is proved that the MEEE coincides with the conditional expectation in case all the densities in the mixture are isotropic and unimodal; further a counter-example is given which shows that the result cannot be generalized for symmetric non-isotropic densities.


62B10, 94A17, 62F10