Knowledge Distillation Applied to Optical Channel Equalization: Solving the Parallelization Problem of Recurrent Connection
Sasipim Srivallapanondh, Pedro J. Freire, Bernhard Spinnler, Nelson Costa, Antonio Napoli, Sergei K. Turitsyn, Jaroslaw E. Prilepsky
arXiv:2212.04569v112 citationsh-index: 61
Originality Incremental advance
AI Analysis
This solves a parallelization problem for optical communication systems, but it is incremental as it adapts an existing technique to a specific domain.
The authors tackled the non-parallelizability of recurrent neural network-based equalizers in optical channel equalization by using knowledge distillation to convert them into a parallelizable feedforward structure, achieving a 38% latency decrease with only a 0.5dB impact on the Q-factor.
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.