Evolving connectionist systems: From neuro-fuzzy-, to spiking- and neuro-genetic

Abstract

This chapter follows the development of a class of neural networks (NN) called evolving connectionist systems (ECOS). The term evolving is used here in its meaning of unfolding, developing, changing, revealing (according to the Oxford dictionary) rather than evolutionary. The latter represents processes related to populations and generations of them. An ECOS is a neural network-based model that evolves its structure and functionality through incremental, adaptive learning and self-organization during its lifetime. In principle, it could be a simple NN or a hybrid connectionist system. The latter is a system based on neural networks that also integrate other computational principles, such as linguistically meaningful explanation features of fuzzy rules, optimization techniques for structure and parameter optimization, quantum-inspired methods, and gene regulatory networks. The chapter includes definitions and examples of ECOS such as: evolving neuro-fuzzy and hybrid systems; evolving spiking neural networks, neurogenetic systems, quantum-inspired systems, which are all discussed from the point of view of the structural and functional development of a connectionist-based model and the knowledge that it represents. Applications for knowledge engineering across domain areas, such as in bioinformatics, brain study, and intelligent machines are presented.