Optoelectronic Memristor for Switchable Reservoir Computing and ANN Computing

The traditional von Neumann architecture exhibits low efficiency in handling large-scale data, inspiring the development of novel neuromorphic computing electronics. Here, we propose an optoelectronic synaptic device with a functional layer of Nb2O5/NiO, which simulates various biological synaptic functions for switchable reservoir computing (RC) and artificial neural network (ANN) computing, including excitatory postsynaptic current (EPSC), paired-pulse facilitation (PPF), the transition from short-term memory (STM) to long-term memory (LTM), learning–forgetting–relearning processes, and 16 reservoir states. Synaptic weights modulated by light pulses were applied to an ANN for MNIST handwritten digit recognition and CIFAR-10 image classification, achieving an accuracy of 90.6% and 88.3%. In addition, the device enabled RC based on MNIST through optical pulse sequences, with an accuracy of 94.8%. These results demonstrate the feasibility of optoelectronic synaptic devices for constructing efficient neuromorphic computing systems and offer insights for future applications in complex neural networks.