Blind Equalization for Coherent Optical Communications Based on Variational Inference

We propose a novel equalization technique for coherent optical communications, which can improve the performance of existing linear equalizers, in particular for probabilistic shaping, and enables powerful neural-network-based equalizers.