Streamlined optical training of large-scale modern deep learning architectures with direct feedback alignment
Open Access
- 15 May 2026
- journal article
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 123 (20)
Abstract
Modern deep learning relies nearly exclusively on dedicated electronic hardware accelerators. Photonic approaches, with low consumption and high operation speed, are increasingly considered for inference but, to date, remain mostly limited to relatively basic tasks. Simultaneously, the problem of training deep and complex neural networks, overwhelmingly performed through backpropagation, remains a significant limitation to the size and, consequently, the performance of current architectures and a major compute and energy bottleneck. Here, we experimentally implement a versatile and scalable training algorithm, called direct feedback alignment, on a hybrid electronic-photonic platform. An optical processing unit performs large-scale random matrix multiplications, which is the central operation of this algorithm. We perform optical training of modern deep learning architectures, including Transformers, with more than 1B parameters, and obtain good performances on language, vision, and diffusion-based generative tasks. We study the scaling of the training time and demonstrate a potential advantage of our hybrid opto-electronic approach for ultra-deep and wide neural networks, thus opening a promising route to sustain the exponential growth of modern artificial intelligence beyond traditional von Neumann approaches.Keywords
All Related Versions
Funding Information
- Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (216600)
This publication has 30 references indexed in Scilit:
- Hundred-layer photonic deep learningNature Communications, 2025
- Physical deep learning with biologically inspired training method: gradient-free approach for physical hardwareNature Communications, 2022
- ClimateBench v1.0: A Benchmark for Data‐Driven Climate ProjectionsJournal of Advances in Modeling Earth Systems, 2022
- Photonic matrix multiplication lights up photonic accelerator and beyondLight: Science & Applications, 2022
- Randomized numerical linear algebra: Foundations and algorithmsActa Numerica, 2020
- Backpropagation and the brainNature Reviews Neuroscience, 2020
- Deep learning with coherent nanophotonic circuitsNature Photonics, 2017
- Random synaptic feedback weights support error backpropagation for deep learningNature Communications, 2016
- Deep learningNature, 2015
- Applications of advances in nonlinear sensitivity analysis2005