Abstract

Artificial synapse is basic electronic components used to prepare neuromorphic hardware with integrated storage and computing, high speed and low energy consumption. This work reports an artificial synapse based on perovskite diodes with learning, memory and computing functions. The artificial synapse demonstrates a mechanism for the dual roles of depletion layer width modulation and ion conduction channel, enabling it to complete the calculation task of neural network and memorize data. Artificial synapses are modeled with various neural networks, including single-layer neural networks, deep neural networks, and convolutional neural networks. The results show that artificial synapses show good recognition accuracy for the MNIST dataset. Especially on the convolutional neural network, only relying on more than 3000 parameters has achieved an accuracy rate of 92.4 %.

Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call