Park, Seongsik Jo, Jeonghee 박종길 정연주 김재욱 이수연 곽준영 김인호 박종극 이경석 황규원 장현재 2024-01-12T02:45:28Z 2024-01-12T02:45:28Z 2023-09-04 2023-07-29 https://pubs.kist.re.kr/handle/201004/76401 https://openreview.net/forum?id=UNFsLcAo2o Deep spiking neural networks (SNNs) are promising neural networks for their model capacity from deep neural network architecture and energy efficiency from SNNs’ operations. To train deep SNNs, recently, spatio-temporal backpropagation (STBP) with surrogate gradient was proposed. Although deep SNNs have been successfully trained with STBP, they cannot fully utilize spike information. In this work, we proposed gradient scaling with local spike information, which is the relation between pre- and post-synaptic spikes. Considering the causality between spikes, we could enhance the training performance of deep SNNs. According to our experiments, we could achieve higher accuracy with lower spikes by adopting the gradient scaling on image classification tasks, such as CIFAR10 and CIFAR100. English JMLR-Journal Machine Learning Research Gradient Scaling on Deep Spiking Neural Networks with Spike-Dependent Local Information Conference 1 40th International Conference on Machine Learning (ICML) 40th International Conference on Machine Learning (ICML) US Hawaii, USA 2023-07-23 Workshop on Localized Learning: Decentralized Model Updates via Non-Global Objectives