Deep nonlinear optical neural networks trained with in situ backpropagation
Logan G. Wright, Tatsuhiro Onodera, Martin M. Stein, Tianyu Wang, Darren T. Schachter, Zoey Hu, Peter L. McMahon · 2021
We experimentally demonstrate an optical neural network whose fundamental building blocks are controllable ultrafast quadratic nonlinear pulse-propagation processes. The network is trained in situ by a new backpropagation algorithm that generalizes to arbitrary physical systems.