Jaynes principle versus entanglement

Abstract

We show, by explicit examples, that the Jaynes inference scheme based on maximization of entropy can produce inseparable states even if there exists a separable state compatible with the measured data. It can lead to problems with processing of entanglement. The difficulty vanishes when one uses inference scheme based on minimization of entanglement.

RevTeX, 6 pages, the paper was rewritten to present the problem possibly clearly