Loss of Scientific Knowledge As a Consequence of Publication Requirements

Abstract

In the process of analyzing modern courses on Bayesian methods in machine learning, we discovered there to be a loss of knowledge on the inconsistency of Bayesian learning of the Dirichlet model. We present a particularly simple presentation of such a failure. The reason for the phenomenon of forgetting known scientific facts is a change in the requirements for publications in scientific journals, i.e., references only to modern publications (not older than 5 years).