Birthday paradox, coupon collectors, caching algorithms and self-organizing search

Abstract

This paper introduces a unified framework for the analysis of a class of random allocation processes that include: (i) the birthday paradox; (ii) the coupon collector problem; (iii) least-recently-used (LRU) caching in memory management systems under the independent reference model; (iv) the move-to-front heuristic of self-organizing search. All analyses are relative to general nonuniform probability distributions. Our approach to these problems comprises two stages. First, the probabilistic phenomena of interest are described by means of regular languages extended by addition of the shuffle product. Next, systematic translation mechanisms are used to derive integral representations for expectations and probability distributions.