Training nonlinear optical neural networks with Scattering Backpropagation

2025-08-15

AI Summary

Nonlinear optical neural networks lack efficient, general-purpose, and physics-model-free gradient training methods. Method: We propose Scattering Backpropagation—a model-agnostic technique that approximates full-parameter gradients using only two scattering experiments. It constructs a gradient estimation framework based on the scattering matrix and incorporates a nonreciprocity bias control mechanism to enable experiment-driven backward propagation approximation, eliminating reliance on mathematical models of nonlinear physical processes. The method is broadly applicable across photonic, microwave, and circuit-based physical platforms. Contribution/Results: Evaluated on XOR and MNIST tasks, it achieves high-accuracy training while significantly reducing energy consumption and computational overhead. This work presents the first experimental realization of model-free gradient measurement for nonlinear optical neural networks, overcoming the deployment bottleneck of conventional backpropagation in physical systems. It demonstrates strong generality and scalability across diverse hardware platforms.

Technology Category

• Search and Optimization: Learning to Search
• Machine Learning: Optimization
• Natural Language Processing: Learning & Optimization for NLP

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

As deep learning applications continue to deploy increasingly large artificial neural networks, the associated high energy demands are creating a need for alternative neuromorphic approaches. Optics and photonics are particularly compelling platforms as they offer high speeds and energy efficiency. Neuromorphic systems based on nonlinear optics promise high expressivity with a minimal number of parameters. However, so far, there is no efficient and generic physics-based training method allowing us to extract gradients for the most general class of nonlinear optical systems. In this work, we present Scattering Backpropagation, an efficient method for experimentally measuring approximated gradients for nonlinear optical neural networks. Remarkably, our approach does not require a mathematical model of the physical nonlinearity, and only involves two scattering experiments to extract all gradient approximations. The estimation precision depends on the deviation from reciprocity. We successfully apply our method to well-known benchmarks such as XOR and MNIST. Scattering Backpropagation is widely applicable to existing state-of-the-art, scalable platforms, such as optics, microwave, and also extends to other physical platforms such as electrical circuits.

Problem

• Efficient gradient extraction for nonlinear optical neural networks
• Physics-based training without mathematical model of nonlinearity
• Applicable to scalable platforms like optics and microwaves

Innovation

• Scattering Backpropagation for gradient measurement
• Model-free nonlinear optical training method
• Two scattering experiments extract gradients