Applying Neural Networks in Optical Communication Systems: Possible Pitfalls

TL;DR

This paper examines potential overestimations of neural network performance in optical communication systems caused by short pattern repetitions, highlighting the risk of artificial gains from pattern prediction rather than true channel compensation.

Contribution

It identifies and demonstrates the pitfalls of neural network application in optical systems, emphasizing the need for careful evaluation to avoid overestimating gains.

Findings

Neural networks can produce artificial performance gains with short repetitive patterns.

Performance gains may not reflect true channel compensation but pattern prediction.

Caution is needed when interpreting neural network results in systems with limited memory or pseudo-random sequences.

Abstract

We investigate the risk of overestimating the performance gain when applying neural network based receivers in systems with pseudo random bit sequences or with limited memory depths, resulting in repeated short patterns. We show that with such sequences, a large artificial gain can be obtained which comes from pattern prediction rather than predicting or compensating the studied channel/phenomena.