Research map / Architectures and efficiency / Spiking networks
Online Training Through Time for Spiking Neural Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Replace matched-setting BPTT with OTTT_A | VGG (sWS) · CIFAR-10 | accuracy: 92.78±0.34% -> 93.52±0.06% | Helped |
| Replace matched-setting BPTT with OTTT_A | VGG (sWS) · CIFAR-100 | accuracy: 69.06±0.07% -> 71.05±0.04% | Helped |
| Replace BPTT with OTTT_A | NF-ResNet-34 · ImageNet | accuracy: 63.72% -> 65.15% | Helped |
| Replace matched-setting BPTT with OTTT_A | VGG (sWS) · DVS-CIFAR10 | accuracy: 72.60±1.26% -> 76.27±0.05% | Helped |
| Replace matched VGG BPTT with OTTT_A | VGG (sWS) · DVS128-Gesture | accuracy: 96.88% -> 96.88% | No clear effect |
| Update weights online at each time step instead of accumulating gradients before updating (ablation) | OTTT on CIFAR-10 · CIFAR-10 | accuracy: 93.52±0.06% -> 93.49±0.17% | No clear effect |
| Update weights online at each time step instead of accumulating gradients before updating (ablation) | OTTT on CIFAR-100 · CIFAR-100 | accuracy: 71.05±0.04% -> 71.05±0.06% | No clear effect |
| Update weights online at each time step instead of accumulating gradients before updating (ablation) | OTTT on ImageNet · ImageNet | accuracy: 65.15% -> 64.16% | Hurt |
| Update weights online at each time step instead of accumulating gradients before updating (ablation) | OTTT on DVS-CIFAR10 · DVS-CIFAR10 | accuracy: 76.27±0.05% -> 76.63±0.34% | Helped |
| Add a feedback connection from the last feature layer to the first feature layer (ablation) | OTTT_O with VGG on CIFAR-100 · CIFAR-100 | accuracy: 71.05±0.06% -> 72.63±0.23% | Helped |
| Train with batch size 1 instead of batch size 128 for 20 epochs (ablation) | OTTT_A with VGG · CIFAR-10 | accuracy: 88.20% -> 88.07% | No clear effect |
| Train with batch size 1 instead of batch size 128 for 20 epochs (ablation) | OTTT_O with VGG · CIFAR-10 | accuracy: 88.62% -> 88.50% | No clear effect |
Closest papers by meaning
- Temporal Efficient Training of Spiking Neural Network via Gradient Re-weighting (ICLR 2022 · similarity 0.93)
- NDOT: Neuronal Dynamics-based Online Training for Spiking Neural Networks (ICML 2024 · similarity 0.93)
- Advancing Training Efficiency of Deep Spiking Neural Networks through Rate-based Backpropagation (NeurIPS 2024 · similarity 0.91)
- Temporal Flexibility in Spiking Neural Networks: Towards Generalization Across Time Steps and Deployment Friendliness (ICLR 2025 · similarity 0.91)
- Enhancing Training of Spiking Neural Network with Stochastic Latency (AAAI 2024 · similarity 0.90)
- Enhanced Self-Distillation Framework for Efficient Spiking Neural Network Training (NeurIPS 2025 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.