ECOS: Evolving Connectionist Systems and the ECO Learning Paradigm

The paper presents a framework called ECOS for Evolving COnnectionist Systems. ECOS evolve through incremental learning. They can accommodate any new input data, including new features, new classes, etc. New connections and new neurons are created during operation. The ECOS framework is used to develop a particular type of evolving neural networks - evolving fuzzy neural networks. A novel training method is introduced and called eco training. ECOS are four (to six) orders of magnitude faster than the multilayer perceptrons, or fuzzy neural networks, trained with the backpropagation algorithm. This is illustrated on benchmark problems, as well as on a real-time problem such as the task of voice recognition and person identification.

• Type: Conference
• Year: 1998
• Where: ICONIP
• Authors: Nikola K. Kasabov
• Keywords: ECOS, Framework Called Ecos, Fuzzy Neural Networks, ICONIP 1998, ICONIP 2007