Optical AI chips have struggled to implement a crucial algorithm used to train neural networks–backpropagation. But in a new paper in the journal Science, a team from Stanford University has described the first ever implementation of the training approach on a photonic chip.
Comments are closed.