Efficiently Training Time-to-First-Spike Spiking Neural Networks from Scratch

Kaiwei Che, Zhengyu Ma, Yifan Huang, Peng Xue, Li Yuan, Wei Fang, Yonghong Tian
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:13316-13328, 2026.

Abstract

Spiking Neural Networks (SNNs), with their event-driven and biologically inspired mechanisms, are well-suited for energy-efficient neuromorphic hardware. Neural coding, which is critical to SNNs, determines how information is represented via spikes. While Time-to-First-Spike (TTFS) coding uses a single spike per neuron to offer extreme sparsity and energy efficiency, it often suffers from unstable training and low accuracy due to its sparse firing. To address these challenges, we propose a training framework that incorporates parameter initialization, training normalization, a temporal output decoder, and a re-evaluation of the pooling layer. The proposed parameter initialization and training normalization mitigate signal diminishing and gradient vanishing, which helps stabilize training. Our output decoder aggregates temporal spikes to encourage earlier firing, thereby reducing latency. The re-evaluation of the pooling layer demonstrates that max-pooling violates single-spike constraints, which should be avoided, whereas average-pooling preserves them. Experiments show that our framework stabilizes and accelerates training, reduces latency, and achieves state-of-the-art accuracy for step-by-step TTFS SNNs on MNIST ($99.48%$), Fashion-MNIST ($92.90%$), CIFAR10 ($90.56%$), CIFAR100 ($70.27%$) and DVS Gesture ($95.83%$).

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-che26a, title = {Efficiently Training Time-to-First-Spike Spiking Neural Networks from Scratch}, author = {Che, Kaiwei and Ma, Zhengyu and Huang, Yifan and Xue, Peng and Yuan, Li and Fang, Wei and Tian, Yonghong}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {13316--13328}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/che26a/che26a.pdf}, url = {https://proceedings.mlr.press/v306/che26a.html}, abstract = {Spiking Neural Networks (SNNs), with their event-driven and biologically inspired mechanisms, are well-suited for energy-efficient neuromorphic hardware. Neural coding, which is critical to SNNs, determines how information is represented via spikes. While Time-to-First-Spike (TTFS) coding uses a single spike per neuron to offer extreme sparsity and energy efficiency, it often suffers from unstable training and low accuracy due to its sparse firing. To address these challenges, we propose a training framework that incorporates parameter initialization, training normalization, a temporal output decoder, and a re-evaluation of the pooling layer. The proposed parameter initialization and training normalization mitigate signal diminishing and gradient vanishing, which helps stabilize training. Our output decoder aggregates temporal spikes to encourage earlier firing, thereby reducing latency. The re-evaluation of the pooling layer demonstrates that max-pooling violates single-spike constraints, which should be avoided, whereas average-pooling preserves them. Experiments show that our framework stabilizes and accelerates training, reduces latency, and achieves state-of-the-art accuracy for step-by-step TTFS SNNs on MNIST ($99.48%$), Fashion-MNIST ($92.90%$), CIFAR10 ($90.56%$), CIFAR100 ($70.27%$) and DVS Gesture ($95.83%$).} }
Endnote
%0 Conference Paper %T Efficiently Training Time-to-First-Spike Spiking Neural Networks from Scratch %A Kaiwei Che %A Zhengyu Ma %A Yifan Huang %A Peng Xue %A Li Yuan %A Wei Fang %A Yonghong Tian %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-che26a %I PMLR %P 13316--13328 %U https://proceedings.mlr.press/v306/che26a.html %V 306 %X Spiking Neural Networks (SNNs), with their event-driven and biologically inspired mechanisms, are well-suited for energy-efficient neuromorphic hardware. Neural coding, which is critical to SNNs, determines how information is represented via spikes. While Time-to-First-Spike (TTFS) coding uses a single spike per neuron to offer extreme sparsity and energy efficiency, it often suffers from unstable training and low accuracy due to its sparse firing. To address these challenges, we propose a training framework that incorporates parameter initialization, training normalization, a temporal output decoder, and a re-evaluation of the pooling layer. The proposed parameter initialization and training normalization mitigate signal diminishing and gradient vanishing, which helps stabilize training. Our output decoder aggregates temporal spikes to encourage earlier firing, thereby reducing latency. The re-evaluation of the pooling layer demonstrates that max-pooling violates single-spike constraints, which should be avoided, whereas average-pooling preserves them. Experiments show that our framework stabilizes and accelerates training, reduces latency, and achieves state-of-the-art accuracy for step-by-step TTFS SNNs on MNIST ($99.48%$), Fashion-MNIST ($92.90%$), CIFAR10 ($90.56%$), CIFAR100 ($70.27%$) and DVS Gesture ($95.83%$).
APA
Che, K., Ma, Z., Huang, Y., Xue, P., Yuan, L., Fang, W. & Tian, Y.. (2026). Efficiently Training Time-to-First-Spike Spiking Neural Networks from Scratch. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:13316-13328 Available from https://proceedings.mlr.press/v306/che26a.html.

Related Material