A decade of Kasabov’s evolving connectionist systems: a review

Abstract

Evolving connectionist Systems (ECoS) are a family of constructive artificial neural network algorithms that were first proposed by Kasabov in 1998, where ‘evolving’ in this context means “changing over time”, rather than evolving through simulated evolution. A decade on, the number of ECoS algorithms, and the problems to which they have been applied, have multiplied. This paper reviews the current state-of-the-art in the field of ECoS networks via a substantial literature review. It reviews (1) the motivations for ECoS, (2) the major ECoS algorithms in use, (3) previously existing constructive algorithms that are similar to ECoS, (4) empirical evaluations of ECoS networks over benchmark data sets, (5 applications of ECoS to real-world problems. The paper ends with some suggestions of future directions of research into ECoS networks