This paper demonstrates a novel approach to recurrent neural networks using the natural spatiotemporal dynamics of light in multimode optical fibers. The system processes temporal data with high energy efficiency by leveraging passive light propagation and inherent optical nonlinearities, achieving competitive performance across diverse temporal and spatiotemporal learning tasks without trainable optical parameters.

This paper introduces a novel hybrid recurrent neural network architecture that integrates linear structures like expert models and Kalman filters for day-ahead electricity price forecasting. The model achieves approximately 12% higher accuracy than leading benchmarks by effectively capturing both linear and nonlinear price characteristics in power markets.