A Compact Behavioral Model Quantifying the Relationship between Optoelectronic Memristor Dynamics and Reservoir Computing Performance

Optoelectronic memristors are promising platforms for neuromorphic and reservoir computing because of their intrinsic nonlinearity and short-term memory dynamics. However, the relationship between experimentally measurable device dynamics and computational capability remains unclear. Here, we present a compact behavioral modeling framework that establishes a direct link between experimentally measured excitatory postsynaptic current (EPSC) dynamics and reservoir computing performance. A parametric exponential model is developed to describe the temporal evolution of memristor current under single- and multipulse stimulation, enabling accurate reconstruction of history-dependent responses using a small set of physically interpretable parameters. By integrating the model into the reservoir layer, a reproducible memristor-driven reservoir computing system is constructed for diverse temporal recognition tasks. Furthermore, bit-depth scaling analysis reveals that higher-resolution dynamic state encoding improves convergence and recognition performance for complex tasks, while excessive encoding resolution yields diminishing returns. This work provides a physically interpretable framework for co-optimizing memristor dynamics, temporal encoding resolution, and reservoir-level computational capability.