ABSTRACT
In Chapters 1 and 2, the HDBT constraints played a major theoretical roˆle. They were responsible for the existence of ML and MAP estimates and for their consistency and robustness. Since the HDBT constant is unknown in practical applications and as there is a consistent free local optimum, Theorem 1.19, we will focus here on algorithms that compute free local optima and steady partitions. They merit the highest attention. Their number may run into the thousands, the great majority being unreasonable. It will, however, be easy to single out the favorite one(s) by the method of scale balance in Section 4.1. Solutions to clustering problems are often not unique.
