Non-negative isomorphic neural networks for photonic neuromorphic accelerators

TL;DR

This paper introduces a method to convert regular neural networks into non-negative isomorphic versions suitable for photonic neuromorphic accelerators, improving hardware compatibility and efficiency while maintaining accuracy.

Contribution

We propose a novel methodology to derive non-negative isomorphic neural networks and a sign-preserving training approach, addressing hardware constraints in photonic neuromorphic systems.

Findings

Achieved accurate non-negative neural network equivalents of regular models.

Enabled training of non-negative networks with sign-preserving optimization.

Facilitated deployment of neural networks on incoherent photonic hardware.

Abstract

Neuromorphic photonic accelerators are becoming increasingly popular, since they can significantly improve computation speed and energy efficiency, leading to femtojoule per MAC efficiency. However, deploying existing DL models on such platforms is not trivial, since a great range of photonic neural network architectures relies on incoherent setups and power addition operational schemes that cannot natively represent negative quantities. This results in additional hardware complexity that increases cost and reduces energy efficiency. To overcome this, we can train non-negative neural networks and potentially exploit the full range of incoherent neuromorphic photonic capabilities. However, existing approaches cannot achieve the same level of accuracy as their regular counterparts, due to training difficulties, as also recent evidence suggests. To this end, we introduce a methodology to obtain the non-negative isomorphic equivalents of regular neural networks that meet requirements of neuromorphic hardware, overcoming the aforementioned limitations. Furthermore, we also introduce a sign-preserving optimization approach that enables training of such isomorphic networks in a non-negative manner.