On-Line Expansion of Output Space in Evolving Fuzzy Neural Networks
Akbar Ghobakhlou, Michael Watts and Nikola Kasabov
The paper presents a methodology for expanding the number of classes an Evolving Fuzzy Neural Network(EFuNN) is able to classify. This is a useful trait for such applications as adaptive speech recognition systems, and strongly complements the ability of EFuNNs to adapt to new examples of known classes. Experiments with isolated word recognition demonstrate the efficacy of EFuNNs in this problem domain, while further experiments expand these networks with new words, demonstrating the methodology that is the crux of this work. The experimental results show that the suggested methodology is a promising approach to the problem of expanding adaptive connectionist classification systems to accommodate new classes.
send email to authors

Last update 29-07-2000, all inquiries to Bernhard Sendhoff