Neural nets filters: integrated coding and signaling in communication systems

Abstract

The authors describe the potential of neural net filters in communication systems. They consider applications of neural networks in those fields associated with communications where time-varying linear systems need to be used; the structure of the neural net considered is the multiple-layer feed-forward network. It is shown that an FIR (finite impulse response) filter with finite representation of its output could be viewed as a two-layer neural net. Experiments on the equalization of nonlinear communication channels with memory are reported, demonstrating the potential of neural networks in integrated tools for signal processing and decoding.<>