High-speed reservoir computing using photonic integrated circuit optical parametric oscillators.

Abstract

Over the past decade, deep learning has led to disruptive advancements with key applications in computer vision, natural language processing, and predictive analytics. With the increasing prevalence and adoption of deep learning algorithms, the quest for hardware solutions that can efficiently process data in real time with high speeds and low latencies has come to the forefront of research. On-chip photonic neural networks offer a promising platform that leverage high bandwidths and low propagation losses associated with light to perform analog deep learning computations. However, nanophotonic circuits are yet to achieve the required linear and nonlinear operations simultaneously in an all-optical and ultrafast fashion. Here, we report a high-speed photonic integrated circuit (PIC)-based reservoir computer using an optical parametric oscillator (OPO) fabricated on thin-film lithium niobate. We apply the PIC-based OPO computing system for a variety of benchmark tasks including chaotic time series prediction, nonlinear error correction in a noisy communication channel, and noisy waveform classification, achieving >93% accuracies at an operating clock rate of ∼10 gigahertz in all cases. Our OPO network can allow for subnanosecond latencies when implemented in an end-to-end all-optical fashion, a timescale that is shorter than a single clock cycle in state-of-the-art digital electronic processors. By circumventing the need for optical-electronic-optical conversions, our high-speed PIC-based reservoir system paves the way for the next generation of compact, energy-efficient, all-optical neural networks with ultralow latencies.