Optical next generation reservoir computing

Artificial neural networks with dynamics exhibit remarkable capability in processing information. Reservoir computing (RC) is a canonical example that features rich computing expressivity and compatibility with physical implementations for enhanced efficiency. Recently, a new RC paradigm known as next generation reservoir computing (NGRC) further improves expressivity but compromises the physical openness, posing challenges for neuromorphic realizations. Here we demonstrate optical NGRC with large-scale computations performed by light scattering through disordered media. In contrast to conventional optical RC implementations, we drive our optical reservoir directly with time-delay inputs. We show that, much like digital NGRC that relies on polynomial features of delayed inputs, our optical reservoir also implicitly generates these polynomial features for desired functionalities. By leveraging the domain knowledge of the reservoir inputs, the optical NGRC not only predicts the short-term dynamics of the low-dimensional Lorenz63 and high-dimensional Kuramoto-Sivashinsky chaotic time series, but also replicates their long-term ergodic properties. Optical NGRC shows superiority in shorter training length, fewer hyperparameters and increased interpretability compared to conventional optical RC, while achieving state-of-the-art forecasting performance. Given its scalability and versatility, the optical NGRC framework also paves the way for next generation physical RC, new applications and architectures in a broad sense.