Knowledge Distillation Applied to Optical Channel Equalization: Solving the Parallelization Problem of Recurrent Connection

TL;DR

This paper introduces a knowledge distillation method to convert RNN-based optical channel equalizers into parallelizable feedforward models, significantly reducing latency with minimal performance loss.

Contribution

The novel approach applies knowledge distillation to enable parallelization of RNN equalizers, addressing a key bottleneck in optical communication systems.

Findings

38% latency reduction achieved

Q-factor decreased by only 0.5dB

Enables efficient parallel processing of equalizers

Abstract

To circumvent the non-parallelizability of recurrent neural network-based equalizers, we propose knowledge distillation to recast the RNN into a parallelizable feedforward structure. The latter shows 38\% latency decrease, while impacting the Q-factor by only 0.5dB.