Machine learning for photonics: from computing to communication

Neural networks are effective tools for learning direct and inverse models. Here, we review two specific applications of neural networks to photonics: (i) learning accurate direct models for optical matrix multipliers and (ii) inverse modeling for short-reach fiber communication systems, enabling signal equalization.

Automated summary

This article reviews two specific applications of neural networks in photonics: learning direct models for optical matrix multipliers and inverse modeling for short-reach fiber communication systems.