上海科技大学知识管理系统

Artificial perception learning systems and artificial neural networks based on neuromorphic devices would promote the progress of neuromorphic engineering to a great extent. Here, we propose a vision enhancement and information decoding platform using aqueous solution-processed mesoporous silica coating gated oxide ionotronic neuromorphic transistors. These transistors exhibit excellent electrical performances with the ability to synergistically respond to both optical and electrical stimuli. Interestingly, an effective linear synaptic weight updating strategy is proposed. Thus, an artificial neural network is built and evaluated through simulations. With the optimized current spiking synaptic weight updating, an excellent recognition accuracy of similar to 94.73% is demonstrated after 125 learning epochs in recognizing the MNIST handwritten digits. The accuracy is comparable to the ideal accuracy of similar to 94.72%. Finally, the information decoding function is conceptually demonstrated with the photoelectric synergic responses. The proposed photoelectric neuromorphic transistors have great potential application in the fields of artificial visual platforms and bionic perception learning systems.