Recently, a research group led by Professor Xu Hui from College of Electronic Science and Technology, National University of Defense Technology collaborated with a research group led by Professor Liu Qi from Fudan University to report a leaky integrate-and-fire (LIF) neuron based on a CMOS-compatible Hf0.2Zr0.8O2 anti-ferroelectric field-effect transistor (AFeFET). The AFeFET neuron features CMOS-compatible, tunable firing frequency, ultra-low hardware cost, ultra-low energy consumption, high endurance, and high uniformity. The research constructs a two-layer spiking neural network (SNN) and tests its autonomous learning ability, paving way for higher-efficient and large-scale neuromorphic computing chips. The research result is published in Nature Communications on November 17, entitled "Compact artificial neuron based on anti-ferroelectric transistor".
According to the research, the designed AFeFET neuron achieves ultra-low hardware cost with no capacitance and additional reset circuit, and thus increased the integration level of neuromorphic computing chips. The two-layer SNN (784 × 400 × 10) achieves 96.8% of recognition accuracy, providing prospects of application in image and speech intelligent processing.
The first authors are Dr. Cao Rongrong from College of Electronic Science and Technology, National University of Defense Technology, Associate Research Fellow Zhang Xumeng from Fudan University, and Associate Professor Liu Sen from National University of Defense Technology. Corresponding authors are Associate Professor Li Qingjiang from National University of Defense Technology and Professor Liu Qi from Fudan University.