Autonomous library for evolutionary algorithms

Abstract

The implementation of an evolutionary algorithm is always tweaked to a certain problem and is therefore not portable to the other problem domains. This is mainly because of the sensitive fitness function that ties the evolution to the problem. In order to create a general EA library, we avoided the regular fitness evaluation and use a more natural implicit fitness concept. We created an independent EA library that needs minimal programming to create solutions for different problem areas. The library engine evolves the two types of individuals, which holds the discovered knowledge, and uses a non-standard implicit fitness evaluation in a co-evolving environment. This library was used to create an EA program for the induction of decision trees.