A FUSION OF FUZZY SETS AND LAYERED NEURAL NETWORKS AT THE INPUT, OUTPUT AND NEURONAL LEVELS

A way of incorporating the concepls of fU2.ZY sets into layered neural networks has been described. The inpul can be provided to quanlitative or linguistic fOnTIS while Ihe outpUI may be modeled as membership values. Logical operators, viz., I-norm T and r-conorm S involving And and Or neurons. are employed al the neuronal level, and the conventional back propagarion algorilhm is accordingly modified using various fuzzy implication operalors. TIle usefulness of Ihe model fOT classification is demonstrated on a set of vowel data by developing various melhods along wilh Iheir comparison. Effects of fuzZificalion al the Input and outpUI are also investigated. 1. INTRODUCIlON Anificial neural networks 1-9 are signal processing systems that try to emulate the human brain, i.e., the behaviour of biological nervous systems, by providing a mathematical model of combination of numerous connected in a network. These models are reputed to have the following characteristics adaptivity (the ability to adjust when given new information), speed (via massive parallelism), fault tolerance (to missing, confusing and/or noisy data) and optimality (as regards error rates I