Integrated diffractive neural networks performing optical inference

Photonics has been proposed as a paradigm shift for deep learning and, in general, machine learning, using photons instead of electrons for computation, since photonic hardware offers high-speed optical communication (speed of light in medium) and massive parallelism of optical signals (multiplexing in time, space, wavelength, polarization, orbital angular momentum, etc.) 1 . In particular, the idea of build artificial neural networks on top of an optical field propagating through free space or some medium has recently attracted a lot of attention, as it promises to pave the way for a future where optical hardware can process data and execute machine learning functionalities at the speed of light 2 .