Brain-like Functions in Evolving Connectionist Systems for On-line, Knowledge-Based Learning

Abstract

The paper discusses some biological principles of the human brain that would be useful to implement in intelligent information systems (IS). These principles are used to formulate seven major requirements to the current and the future IS. These requirements are met in a new connectionist architecture called evolving connectionist systems (ECOS). ECOS are designed to facilitate building on-line, adaptive, knowledge-based IS. ECOS evolve through incremental, hybrid (supervised/unsupervised), on-line learning. They can accommodate new input data, including new features, new classes, etc. through local element tuning. The ECOS framework is presented and illustrated on a particular type of evolving neural networks evolving fuzzy neural networks (EFuNNs). EFuNNs can learn spatial-temporal sequences in an adaptive way, through one pass learning. Rules can be inserted and extracted at any time of the system operation. ECOS and EFuNNs are suitable for adaptive pattern classification; adaptive, phonemebased spoken language recognition; adaptive dynamic time-series prediction; intelligent agents.