Genetic Algorithms for the Design of Fuzzy Neural Networks

The paper presents a methodology for designing the structure of a fuzzy neural network in a multi-modular connectionist system for classification purposes and illustrates the methodology on the task of phoneme recognition of the 43 phonemes in New Zealand English. The results show that by using this methodology the recognition rate can be improved significantly when compared to the recognition rate of the same modules designed manually.

• Type: Conference
• Year: 1998
• Where: ICONIP
• Authors: Michael J. Watts, Nikola K. Kasabov
• Keywords: Fuzzy Neural Network, ICONIP 1998, ICONIP 2007, Methodology, Recognition Rate