Dual adaptive training of photonic neural networks

TL;DR

This paper introduces dual adaptive training (DAT) for photonic neural networks, enabling them to adapt to systematic errors and maintain high performance in large-scale physical implementations, surpassing existing training methods.

Contribution

The paper proposes a novel dual adaptive training method that predicts systematic errors and optimizes model accuracy during deployment of photonic neural networks.

Findings

DAT achieves high similarity mapping between models and physical systems.

DAT maintains classification accuracy in large-scale PNNs with systematic errors.

DAT outperforms state-of-the-art in situ training approaches.

Abstract

Photonic neural network (PNN) is a remarkable analog artificial intelligence (AI) accelerator that computes with photons instead of electrons to feature low latency, high energy efficiency, and high parallelism. However, the existing training approaches cannot address the extensive accumulation of systematic errors in large-scale PNNs, resulting in a significant decrease in model performance in physical systems. Here, we propose dual adaptive training (DAT) that allows the PNN model to adapt to substantial systematic errors and preserves its performance during the deployment. By introducing the systematic error prediction networks with task-similarity joint optimization, DAT achieves the high similarity mapping between the PNN numerical models and physical systems and high-accurate gradient calculations during the dual backpropagation training. We validated the effectiveness of DAT by using diffractive PNNs and interference-based PNNs on image classification tasks. DAT successfully trained large-scale PNNs under major systematic errors and preserved the model classification accuracies comparable to error-free systems. The results further demonstrated its superior performance over the state-of-the-art in situ training approaches. DAT provides critical support for constructing large-scale PNNs to achieve advanced architectures and can be generalized to other types of AI systems with analog computing errors.