Bloat: Past, Present, Future

Authors

Abstract

Bloat can be defined as an excess of code growth without a corresponding improvement in fitness. This problem has been one of the most intensively studied subjects since the beginnings of Genetic Programming. This report presents a deep analysis of the past and present contributions dealing with the problem of bloat, and discusses the prospects of future challenges and developments. In particular, it reviews all the main theories explaining bloat, and many of the bloat control methods available so far, with particular attention to the new Crossover Bias theory and the bloat control method it inspired, Operator Equalisation. Although still recent and requiring improvements, Operator Equalisation has already proven to be more than just a bloat control method. It reveals novel evolutionary dynamics that allow a successful search without code growth, and shows great potential to be extended and integrated into several different elements of the evolutionary process. The results presented clearly show that Genetic Programming using Operator Equalisation is essentially bloat free. After so many years of research, the true reasons for bloat are finally understood, and the path to its full extinction is clearly seen.

Keywords

Genetic Programming, Bloat, Operator Equalisation, Review

Subject

Genetic Programming