Nonlinear classification and adaptive structures

The main purpose of this paper is to examine a number of possible architectures for nonlinearadaptive filtering, specifically related to adaptive equalisation. The approach taken proceeds by first reformulating the filtering process as a form of classification task in N dimensions. In the case of filtering the dimensionality is determined by the number of data samples in the filter data input vector. The task of classification then proceeds using a number of possible strategies, i.e. the multilayer perceptron, Volterra series modelling and cluster analysis. The techniques are evaluated in comparison with normal linear equalisation procedures