Task-adaptive physical reservoir computing

Abstract

Reservoir computing is a neuromorphic architecture that may offer viable solutions to the growing energy costs of machine learning. In software-based machine learning, computing performance can be readily reconfigured to suit different computational tasks by tuning hyperparameters. This critical functionality is missing in ‘physical’ reservoir computing schemes that exploit nonlinear and history-dependent responses of physical systems for data processing. Here we overcome this issue with a ‘task-adaptive’ approach to physical reservoir computing. By leveraging a thermodynamical phase space to reconfigure key reservoir properties, we optimize computational performance across a diverse task set. We use the spin-wave spectra of the chiral magnet Cu2OSeO3 that hosts skyrmion, conical and helical magnetic phases, providing on-demand access to different computational reservoir responses. The task-adaptive approach is applicable to a wide variety of physical systems, which we show in other chiral magnets via above (and near) room-temperature demonstrations in Co8.5Zn8.5Mn3 (and FeGe).

Physical Reservoir Computing Enabled by Solitary Waves and Biologically Inspired Nonlinear Transformation of Input Data

“…While the biophysical origin of the restoration of the missing fundamental continues to be a subject of debate, it has been suggested that it can be explained by nonlinear and chaotic effects [17]. Nonlinear processes in biological neural systems have also motivated research on artificial neural networks that exploit the nonlinear properties of diverse mathematical models and physical systems [18][19][20][21][22][23][24][25][26][27][28][29].…”

Exploiting Signal Propagation Delays to Match Task Memory Requirements in Reservoir Computing

“…Although software simulated reservoirs offer enough flexibility to implement such task-dependent adaptive rules, doing the same in physical RC poses a significant challenge. However, recently adaptive reservoirs have also been implemented in physical systems by [ 36 ]. By adapting the thermodynamical phase space of the physical substrate, the same reservoir could be optimized for multiple tasks.…”

Classical and Quantum Physical Reservoir Computing for Onboard Artificial Intelligence Systems: A Perspective

“…Importantly, similarly to the efficiency of the brain of some insects, which has a low number of neurons compared with a human brain, RC systems require just several thousands of neurons to undertake certain tasks more efficiently than a high-performance workstation computer running sophisticated software [36,38]. This property is ideally suitable for the development of AI systems for mobile platforms, where the control unit must consume low power while delivering practicable machine learning, vision and sensing capabilities in a real-time regime [71].…”