Proceedings of the International Workshop on Applications of Neural Networks to Telecommunications

Neural Networks for Adaptive Equalization

Timothy X Brown

Bell Communications Research445 South Street, Morristown, NJ [e-mail]

Abstract

Deficiencies in linear approaches to equalization have led to non-linear neural network equalization. Because of the potential for efficient analog hardware implementation, neural networks are especially attractive. A review of the neural literature reveals that indeed neural networks can have improved performance, but at the expense of much longer training times. In this paper we present the results of this review as well as methods for counteracting the recognized deficiencies.

1  Introduction

Within a communications system, equalization is the task of compensating for distorted received ...