On the Algorithmic Complexity, Universal Prior and Ockham's Razor
Nov 30 1991
The first part of this paper is a review of basic notions and results connected with Kolmogorov complexity theory. A few original results are presented in sections 3 and 4; they are not of a statistical nature. emphasis is given to the so called universal prior. Though the prior itself is not a calculable measure it has highly interesting properties from the Bayesian viewpoint.
In the second part of the paper we discuss the principles that emerge from algorithmic complexity theory in the context of statistical prediction and estimation. It is argued that as a rule, the principles re Bayesian in nature.