Optoacoustic Building Blocks for Photonic Neural Networks

The realization of massive machine learning models has integrated artificial intelligence into our daily lives, yet, their training, maintenance, and scalability rely on power-hungry electronic processors [1]. Neuromorphic computing seeks to address these challenges by mimicking neural architechtures directly in physical systems. Photonics is a promising platform as it offers high bandwidth, rapid processing speeds, and low energy consumption [2], [3]. Examples showing the potential of photonics include reservoir computing and extreme learning machines using optical components [4], deep diffractive neural networks [5], and Mach-Zender interferometer meshes [6].