|
An Energy-efficient Adaptive Exponential Integrate and Fire Neuron based on Novel Area-Efficient Fast CORDIC for Spiking Neural Networks
点击次数：
第一作者：Y. Li
通讯作者：C. Wang
合写作者：G. Hong, Z. Peng, J. Wang, J. Xu, G. Yu, C. Huang
发表刊物：IEEE International Conference on Integrated Circuits, Technologies and Applications (ICTA 2021)
收录刊物：EI
文献类型：C
DOI码：10.1109/ICTA53157.2021.9661975
摘要：The realization of bionic neural models has great challenges on the biological characteristics and hardware efficiency of neurons. This paper proposes an energy efficient Adaptive Exponential Integrate and Fire (AdEx) neuron based on a novel Area-Efficient Fast COordinate Rotation Digital Computer (AEF-CORDIC). The novel AEF-CORDIC has a higher energy efficiency and shorter latency by reducing the iteration times with negligible hardware overhead. FPGA implementation results show that the proposed AEF-CORDIC based AdEx Neuron has good biological characteristics. Compared with conventional CORDIC based design, our design has 48.4% and 52.5% improvement on energy efficiency and processing speed, respectively.