Reservoir computing on nanophotonic chips

Abstract

Reservoir computing is a decade old framework from the field of machine learning to use and train recurrent neural networks and it splits the network in a reservoir that does the computation and a simple readout function. This technique has been among the state-of-the- art for a broad class of classification and recognition problems such as time series prediction, speech recognition and robot control. However, so far implementations have been mainly software based, while a hardware implementation offers the promise of being low- power and fast. Despite essential differences between classical software implementation and a network of semiconductor optical amplifiers, we will show that photonic reservoirs can achieve an even better performance on a benchmark isolated digit recognition task, if the interconnection delay is optimized and the phase can be controlled. In this paper we will discuss the essential parameters needed to create an optimal photonic reservoir designed for a certain task. This design can lead to an efficient implementation of a photonic reservoir on a nanophotonic chip.