Reservoir computing has emerged as a promising machine-learning approach to prediction and control of complex nonlinear dynamical systems, rendering it important to explore schemes of physical realization. We articulate two frameworks of physical reservoir computing based on the electrophysiological mechanisms in mammalian neuronal networks. The first emulates sensory-motor coordination triggered by external stimuli, while the second mirrors modulatory inputs that regulate the neural state transitions. Both frameworks utilize a simplified yet dynamically rich, map-based behavioral neural model that preserves the essential neuronal functionalities. Computations conducted with sparse random interconnected networks and uncoupled topologies establish the workings of the proposed frameworks in terms of training, validation, and testing. These findings underline the potential of the proposed frameworks as foundational models for actual physical implementation of reservoir computing.
© 2025 Author(s). Published under an exclusive license by AIP Publishing.