Abstract
Physical reservoir computing (RC) utilizes intrinsic dynamics of physical systems for efficient data processing. Emerging optoelectronic RC platforms merge the benefits of electronic and photonic computation. However, conventional designs are constrained by the unipolar photoresponse of optoelectronic devices, limiting reservoir state diversity and computational accuracy. Here we present an all-optically controlled RC system employing highly uniform and stable oxide memristor arrays. The devices exhibit wavelength-dependent bipolar photoresponse, originating from light-induced dynamic evolution of oxygen vacancies. Tuning the power density and irradiation mode of dual-wavelength light enables dynamic control of photocurrent relaxation and nonlinearity. We thus develop bipolar and parallel coding strategies to significantly enhance reservoir dynamics and nonlinear mapping capability. In word recognition and time-series prediction tasks, the bipolar coding demonstrates markedly improved accuracy compared to unipolar coding. The parallel coding supports multi-source signal fusion within a single reservoir, maintaining high accuracy while significantly reducing hardware consumption. This work provides a high-performance physical RC scheme, paving the way for intelligent edge computing.