Physical Reservoir Computing with Over-Moded Complex Systems

The ceaseless development of machine learning technologies has introduced high-performance computation of a wide range of tasks, although at the expense of larger computational costs. The execution of machine learning (ML) algorithms largely depends on its computing `substrate', which is often not optimized for running ML tasks. Thus, the investigation of alternative computing methods with tailored physical structures has attracted a great deal of research interest.

Here we utilize the wave dynamics inside a chaotic-shaped electromagnetic cavity containing nonlinear elements to emulate the complex dynamics of reservoir computing (RC), which is a genre of recurrent neural network (RNN). In the short-wavelength limit, the evolution of electromagnetic fields inside chaotic-shaped enclosures shows complex patterns, and is extremely sensitive to perturbations.  We propose unique techniques to create virtual RC nodes by both frequency stirring and spatial perturbation. The computational power of the wave chaotic RC is demonstrated with various benchmark tasks. The wave chaos RC is also mechanically robust, and can be scaled to a wide range of wavelengths from RF to the visible.