Feedforward networks with fuzzy signals

Abstract The paper discusses feedforward neural networks with fuzzy signals. We analyze the feedforward phase and show some properties of the output function. Then we present a backpropagation like adaptation algorithm for crisp weights, thresholds and neuron slopes of the multilayer network with sigmoidal transfer functions. We provide theoretical justification for the adaptation formulas. The results are of general nature and together with the presented approach can be used for other types of feedforward networks. Proposed and discussed are also applications of the presented feedforward networks.