LuxIA: A Lightweight Unitary matriX-based Framework Built on an Iterative Algorithm for Photonic Neural Network Training

2025-12-24

AI Summary

To address the prohibitive memory and computational overhead—as well as poor scalability—arising from transmission matrix computation in large-scale photonic neural network (PNN) training, this paper proposes an iterative slicing-based lightweight unitary matrix computation framework. The method innovatively decomposes large-scale unitary matrices into small, iteratively updated sub-blocks, enabling end-to-end backpropagation while ensuring linear memory growth and efficient gradient computation. By integrating explicit unitary constraint modeling with parameterized photonic circuit simulation, the framework enables, for the first time, full-stack training of PNNs comprising thousands of optical units. Experiments on MNIST, Digits, and Olivetti Faces demonstrate a 3.2–8.7× speedup in training and a reduction in memory consumption to one-fifth of prior approaches, thereby substantially overcoming key scalability bottlenecks in PNN training.

Technology Category

• Machine Learning: Matrix & Tensor Methods
• Computer Vision: Learning & Optimization for CV
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

PNNs present promising opportunities for accelerating machine learning by leveraging the unique benefits of photonic circuits. However, current state of the art PNN simulation tools face significant scalability challenges when training large-scale PNNs, due to the computational demands of transfer matrix calculations, resulting in high memory and time consumption. To overcome these limitations, we introduce the Slicing method, an efficient transfer matrix computation approach compatible with back-propagation. We integrate this method into LuxIA, a unified simulation and training framework. The Slicing method substantially reduces memory usage and execution time, enabling scalable simulation and training of large PNNs. Experimental evaluations across various photonic architectures and standard datasets, including MNIST, Digits, and Olivetti Faces, show that LuxIA consistently surpasses existing tools in speed and scalability. Our results advance the state of the art in PNN simulation, making it feasible to explore and optimize larger, more complex architectures. By addressing key computational bottlenecks, LuxIA facilitates broader adoption and accelerates innovation in AI hardware through photonic technologies. This work paves the way for more efficient and scalable photonic neural network research and development.

Problem

• Reduces memory and time for large-scale photonic neural network training
• Enables scalable simulation of complex photonic architectures efficiently
• Overcomes computational bottlenecks in photonic neural network simulation tools

Innovation

• Slicing method reduces memory and time for matrix calculations
• LuxIA framework integrates Slicing for scalable photonic network training
• Experimental results show LuxIA outperforms existing tools in speed