Learning in networks: Complex-valued neurons, pruning, and rule extraction

Abstract

This paper focuses on neural networks with complex-valued (CV) neurons as well as on selected aspects of neural networks learning, pruning and rule extraction. CV neurons can be used as versatile substitutes in real-valued perceptron networks. Learning of CV layers is discussed in context of traditional multilayer feedforward architecture. Such learning is derivative-free and it usually requires networks of reduced size. Selected examples and applications of CV-networks in bioinformatics and pattern recognition are discussed. The paper also covers specialized learning techniques for logic rule extraction. Such techniques include learning with pruning, and can be used in expert systems, and other applications that rely on models developed to fit measured data.