Brain-Inspired Multi-Time Scale Reservoir Computing with Organic Optoelectronic Memristors

Inspired by the multitime scale parallel processing mechanism of the human brain, an organic optoelectronic memristor is designed and fabricated in this work, which exhibits multiple wavelength-dependent temporal relaxation behaviors under different wavelengths of light. Based on this, a hybrid reservoir computing system is constructed. Without increasing the number of physical nodes, the system enables parallel feature extraction of multifrequency components from input signals. In the UCI HAR recognition task, by integrating multiscale dynamics, the recognition accuracy is improved from the maximum of 77.5% achieved with a single-relaxation reservoir to 86.1%. This work provides a feasible bioinspired pathway to overcome the single-time scale bottleneck─which limits feature diversity and representational capacity in reservoir computing─via intrinsic multitime scale dynamics, opening new prospects for developing high-performance and energy-efficient neuromorphic hardware.