Scaling Laws and Topology-Properties of Boolean Reinforcement Learning in Photonic Neural Networks

Abstract

Since recent years artificial intelligence and more particularly neural networks play a major role in our technological societies. Nevertheless, neural networks still remain emulated by traditional computers, resulting in challenging problems such as parallelization, energy efficiency and potentially speed. A change of paradigm is desirable but implementating neural networks in hardware is a non-trivial challenge. One highly promising avenue are optical neural networks [1], potentially avoiding parallelization bottlenecks.