FAST ADAPTIVE DIGITAL EQUALIZATION BY RECURRENT NEURAL NETWORKS

In recent years, neural networks (NN's) have been extensively applied to many signal processing problems. In particular, due to their capaci ty to form complex decision regions, NN's have been successfully used in adaptive equalization of digital communication channels. The mean s quare error (MSE) criterion, which is usually adopted in neural learni ng, is not directly related to the minimization of the classification error, i.e., bit error rate (BER), which is of interest in channel equ alization. Moreover, common gradient-based learning techniques are oft en characterized by slow speed of convergence and numerical ill condit ioning, In this paper, we introduce a novel approach to learning in re current neural networks (RNN's) that exploits the principle of discrim inative learning, minimizing an error functional that is a direct meas ure of the classification error. The proposed method extends to RNN's a technique applied with success to fast learning of feedforward NN's and is based on the descent of the error functional in the space of th e linear combinations of the neurons (the neuron space); its main feat ures are higher speed of convergence and better numerical conditioning w.r.t. gradient-based approaches. whereas numerical stability is assu red by the use of robust least squares solvers. Experiments regarding the equalization of PARI signals in different transmission channels ar e described, which demonstrate the effectiveness of the proposed appro ach.