Light-in-the-loop: using a photonics co-processor for scalable training of neural networks

Abstract

As neural networks grow larger and more complex and data-hungry, training costs are skyrocketing. Especially when lifelong learning is necessary, such as in recommender systems or self-driving cars, this might soon become unsustainable. In this study, we present the first optical co-processor able to accelerate the training phase of digitally-implemented neural networks. We rely on direct feedback alignment as an alternative to backpropagation, and perform the error projection step optically. Leveraging the optical random projections delivered by our co-processor, we demonstrate its use to train a neural network for handwritten digits recognition.

2 pages, 1 figure

References in corpus (5)

• Adam: A Method for Stochastic Optimization
• Experiments with Random Projection
• Random Projections through multiple optical scattering: Approximating kernels at the speed of light
• Dissecting the Graphcore IPU Architecture via Microbenchmarking
• Backpropagation through nonlinear units for all-optical training of neural networks

Cited by in corpus (2)

• Resampling the transmission matrix in an aberration-corrected Bessel mode basis
• Direct Feedback Alignment Scales to Modern Deep Learning Tasks and Architectures