Greedy Boolean Learning in Photonic Recurrent Neural Networks
Abstract
We have recently succeeded in the implementation of a large scale recurrent photonic neural network hosting up to 2025 photonic neurons. All net-work internal and readout connections are physically implemented with fully par-allel technology. Based on a digital micro-mirror array, we can train the Boolean readout weights using a greedy version of reinforcement learning. We find that the learning excellently converges. Furthermore, it appears to possess a conven-iently convex-like cost-function and demonstrates exceptional scalability of the learning effort with system size.