Using photonic reservoirs as preprocessors for deep neural networks

Artificial neural networks are very time consuming and energy intensive to train, especially when increasing the size of network in an attempt improve performance. In this paper, we propose preprocess input data a deep using reservoir, which has originally been introduced framework reservoir computing. The key idea paper is use such transform into state higher dimensional state-space, allows process with improved We focus on photonic reservoirs because their fast computation times low-energy consumption. Based numerical simulations delay-based semiconductor laser, show that preprocessed results performance networks. Furthermore, do not need carefully fine-tune parameters preprocessing reservoir.