eduzhai > Applied Sciences > Engineering >

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

  • Save

... pages left unread,continue reading

Document pages: 2 pages

Abstract: As neural networks grow larger and more complex and data-hungry, trainingcosts are skyrocketing. Especially when lifelong learning is necessary, such asin recommender systems or self-driving cars, this might soon becomeunsustainable. In this study, we present the first optical co-processor able toaccelerate the training phase of digitally-implemented neural networks. We relyon direct feedback alignment as an alternative to backpropagation, and performthe error projection step optically. Leveraging the optical random projectionsdelivered by our co-processor, we demonstrate its use to train a neural networkfor handwritten digits recognition.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×