Based On Neural Networks, Fuzzy Inference

Since Zadeh introduced the compositional rule of fuzzy inference ( CRI) in 1973, various methods have been proposed to improve it. This paper applies a new fuzzy neuron to constructed fuzzy neural networks, and they are applied to fuzzy inference. This new method is generalized form of CRI.In [17], a new fuzzy neuron model was presented based on weak T-norm cluster, which can realize many continuous logic operations. Consisting of the new neuron, three fuzzy neural networks are constructed in this paper, they are applied to fuzzy inference, multiple conditional fuzzy inference and multidimensional fuzzy inference respectively. In the neural network corresponding to multidimensional fuzzy inference, the weights are adjusted by means of solution to fuzzy relation equation. Based on these neural networks, the new method of fuzzy inference overcomes a few shortcoming of the conventional CRI, and it is much easier to satisfy consistency principle of fuzzy inference than CRI. Analyzed the properties of the new method, we discovered that it is continuous and monotonic. The paper implements fuzzy inference engines with the new method, the reasoning results prove better performance obtained than CRI. The new method proposed here is a powerful inference tool for fuzzy control systems etc.