Temporal Flexibility in Spiking Neural Networks: Towards Generalization Across Time Steps and Deployment Friendliness
Not on the map. Why?
Abstract (arXiv)
Spiking Neural Networks (SNNs), models inspired by neural mechanisms in the brain, allow for energy-efficient implementation on neuromorphic hardware. However, SNNs trained with current direct training approaches are constrained to a specific time step. This "temporal inflexibility" 1) hinders SNNs' deployment on time-step-free fully event-driven chips and 2) prevents energy-performance balance based on dynamic inference time steps. In this study, we first explore the feasibility of training SNNs that generalize across different time steps. We then introduce Mixed Time-step Training (MTT), a novel method that improves the temporal flexibility of SNNs, making SNNs adaptive to diverse temporal structures. During each iteration of MTT, random time steps are assigned to different SNN stages, with spikes transmitted between stages via communication modules. After training, the weights are deployed and evaluated on both time-stepped and fully event-driven platforms. Experimental results show that models trained by MTT gain remarkable temporal flexibility, friendliness for both event-driven and clock-driven deployment (nearly lossless on N-MNIST and 10.1% higher than standard methods on CIFAR10-DVS), enhanced network generalization, and near SOTA performance. To the best of our knowledge, this is the first work to report the results of large-scale SNN deployment on fully event-driven scenarios.
Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Builds on 40 of its 58 references that are in the map. First posted on arXiv 18 Mar 2025.
Builds on
40 of 58 references.
References · 40
- 2015Deep Residual Learning for Image Recognitioncited by 52,096
- 2015Distilling the Knowledge in a Neural Networkcited by 10,324
- 2017Improved Regularization of Convolutional Neural Networks with Cutoutcited by 1,628
- 2017Visualizing the Loss Landscape of Neural Netscited by 1,130
- 2016XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networkscited by 849
- 2018AutoAugment: Learning Augmentation Policies from Datacited by 730
- 2017Channel Pruning for Accelerating Very Deep Neural Networkscited by 702
- 2017Spatio-Temporal Backpropagation for Training High-performance Spiking Neural Networkscited by 406
- 2021Deep Residual Learning in Spiking Neural Networkscited by 331
- 2015Converting Static Image Datasets to Spiking Neuromorphic Datasets Using Saccadescited by 314
- 2020Going Deeper With Directly-Trained Larger Spiking Neural Networkscited by 278
- 2022Spikformer: When Spiking Neural Network Meets Transformercited by 271
28 more
- 2018SLAYER: Spike Layer Error Reassignment in Timecited by 266
- 2018Going Deeper in Spiking Neural Networks: VGG and Residual Architecturescited by 245
- 2018Direct Training for Spiking Neural Networks: Faster, Larger, Bettercited by 233
- 2019Universally Slimmable Networks and Improved Training Techniquescited by 221
- 2022Temporal Efficient Training of Spiking Neural Network via Gradient Re-weightingcited by 198
- 2023Spike-driven Transformercited by 146
- 2020RMP-SNN: Residual Membrane Potential Neuron for Enabling Deeper High-Accuracy and Low-Latency Spiking Neural Networkcited by 145
- 2021Optimal Conversion of Conventional Artificial Neural Networks to Spiking Neural Networkscited by 139
- 2019The Heidelberg spiking datasets for the systematic evaluation of spiking neural networkscited by 137
- 2023Optimal ANN-SNN Conversion for High-accuracy and Ultra-low-latency Spiking Neural Networkscited by 122
- 2021A Free Lunch From ANN: Towards Efficient, Accurate Spiking Neural Networks Calibrationcited by 117
- 2022Attention Spiking Neural Networkscited by 107
- 2024Spike-driven Transformer V2: Meta Spiking Neural Network Architecture Inspiring the Design of Next-generation Neuromorphic Chipscited by 103
- 2020DIET-SNN: Direct Input Encoding With Leakage and Threshold Optimization in Deep Spiking Neural Networkscited by 71
- 2022GLIF: A Unified Gated Leaky Integrate-and-Fire Neuron for Spiking Neural Networkscited by 63
- 2023Learning Delays in Spiking Neural Networks using Dilated Convolutions with Learnable Spacingscited by 50
- 2022Online Training Through Time for Spiking Neural Networkscited by 49
- 2023SEENN: Towards Temporal Spiking Early-Exit Neural Networkscited by 41
- 2022TCJA-SNN: Temporal-Channel Joint Attention for Spiking Neural Networkscited by 35
- 2016Theory and Tools for the Conversion of Analog to Spiking Convolutional Neural Networkscited by 31
- 2024Spikformer V2: Join the High Accuracy Club on ImageNet with an SNN Ticketcited by 30
- 2023Training Full Spike Neural Networks via Auxiliary Accumulation Pathwaycited by 26
- 2023DYNAP-SE2: a scalable multi-core dynamic neuromorphic asynchronous spiking neural network processorcited by 21
- 2023Unleashing the Potential of Spiking Neural Networks by Dynamic Confidencecited by 15
- 2023Input-Aware Dynamic Timestep Spiking Neural Networks for Efficient In-Memory Computingcited by 11
- 2023SparseProp: Efficient Event-Based Simulation and Training of Sparse Recurrent Spiking Neural Networkscited by 6
- 2024Exploring the Limitations of Layer Synchronization in Spiking Neural Networkscited by 3
- 2022Training Robust Spiking Neural Networks on Neuromorphic Data with Spatiotemporal Fragmentscited by 2
Built on by
Citing papers per year: 2025: 2.
2 papers.