Development of Fuzzy Neural Networks

In this paper, we explain how multi-layer feedforward neural networks can be fuzzified by using fuzzy numbers for inputs, targets and connection weights. First we briefly review a standard three-layer feedforward neural network and its back-propagation learning algorithm. Next we fuzzify the neural network by extending its inputs, targets and connection weights to fuzzy numbers. The input-output relation of each unit of the fuzzified neural network is defined by the extension principle of Zadeh. We also describe how a learning algorithm of the fuzzified neural network can be derived in a similar manner to the back-propagation algorithm. Finally, we illustrate possible application areas of the fuzzified neural network by simple examples.