{"title":"IM-LIF: Improved Neuronal Dynamics With Attention Mechanism for Direct Training Deep Spiking Neural Network","authors":"Shuang Lian;Jiangrong Shen;Ziming Wang;Huajin Tang","doi":"10.1109/TETCI.2024.3359539","DOIUrl":null,"url":null,"abstract":"Spiking neural networks (SNNs) are increasingly applied to deep architectures. Recent works are developed to apply spatio-temporal backpropagation to directly train deep SNNs. But the binary and non-differentiable properties of spike activities force directly trained SNNs to suffer from serious gradient vanishing. In this paper, we first analyze the cause of the gradient vanishing problem and identify that the gradients mostly backpropagate along the synaptic currents. Based on that, we modify the synaptic current equation of leaky-integrate-fire neuron model and propose the improved LIF (IM-LIF) neuron model on the basis of the temporal-wise attention mechanism. We utilize the temporal-wise attention mechanism to selectively establish the connection between the current and historical response values, which can empirically enable the neuronal states to update resilient to the gradient vanishing problem. Furthermore, to capture the neuronal dynamics embedded in the output incorporating the IM-LIF model, we present a new temporal loss function to constrain the output of the network close to the target distribution. The proposed new temporal loss function could not only act as a regularizer to eliminate output outliers, but also assign the network loss credit to the voltage at a specific time point. Then we modify the ResNet and VGG architecture based on the IM-LIF model to build deep SNNs. We evaluate our work on image datasets and neuromorphic datasets. Experimental results and analysis show that our method can help build deep SNNs with competitive performance in both accuracy and latency, including 95.66% on CIFAR-10, 77.42% on CIFAR-100, 55.37% on Tiny-ImageNet, 97.33% on DVS-Gesture, and 80.50% on CIFAR-DVS with very few timesteps.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":null,"pages":null},"PeriodicalIF":5.3000,"publicationDate":"2024-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Transactions on Emerging Topics in Computational Intelligence","FirstCategoryId":"94","ListUrlMain":"https://ieeexplore.ieee.org/document/10433858/","RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
引用次数: 0
Abstract
Spiking neural networks (SNNs) are increasingly applied to deep architectures. Recent works are developed to apply spatio-temporal backpropagation to directly train deep SNNs. But the binary and non-differentiable properties of spike activities force directly trained SNNs to suffer from serious gradient vanishing. In this paper, we first analyze the cause of the gradient vanishing problem and identify that the gradients mostly backpropagate along the synaptic currents. Based on that, we modify the synaptic current equation of leaky-integrate-fire neuron model and propose the improved LIF (IM-LIF) neuron model on the basis of the temporal-wise attention mechanism. We utilize the temporal-wise attention mechanism to selectively establish the connection between the current and historical response values, which can empirically enable the neuronal states to update resilient to the gradient vanishing problem. Furthermore, to capture the neuronal dynamics embedded in the output incorporating the IM-LIF model, we present a new temporal loss function to constrain the output of the network close to the target distribution. The proposed new temporal loss function could not only act as a regularizer to eliminate output outliers, but also assign the network loss credit to the voltage at a specific time point. Then we modify the ResNet and VGG architecture based on the IM-LIF model to build deep SNNs. We evaluate our work on image datasets and neuromorphic datasets. Experimental results and analysis show that our method can help build deep SNNs with competitive performance in both accuracy and latency, including 95.66% on CIFAR-10, 77.42% on CIFAR-100, 55.37% on Tiny-ImageNet, 97.33% on DVS-Gesture, and 80.50% on CIFAR-DVS with very few timesteps.
期刊介绍:
The IEEE Transactions on Emerging Topics in Computational Intelligence (TETCI) publishes original articles on emerging aspects of computational intelligence, including theory, applications, and surveys.
TETCI is an electronics only publication. TETCI publishes six issues per year.
Authors are encouraged to submit manuscripts in any emerging topic in computational intelligence, especially nature-inspired computing topics not covered by other IEEE Computational Intelligence Society journals. A few such illustrative examples are glial cell networks, computational neuroscience, Brain Computer Interface, ambient intelligence, non-fuzzy computing with words, artificial life, cultural learning, artificial endocrine networks, social reasoning, artificial hormone networks, computational intelligence for the IoT and Smart-X technologies.