Research map / Architectures and efficiency / Spiking networks
Towards Training-Free and Accurate ANN-to-SNN Conversion via Activation-Aware Redistribution
Honglin Cao, Shuai Wang, Zijian Zhou, Ammar Belatreche, Wenjie Wei, Yu Liang, Yu Yang, Rui Xi, Malu Zhang, Haizhou Li
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add channel-wise threshold (ablation) | layer-wise IF baseline, ViT-L · ImageNet-1K T=64 | top-1 accuracy: 73.17 -> 84.27 | Improved |
| add channel-wise threshold (ablation) | layer-wise IF baseline, ViT-L · ImageNet-1K T=8 | top-1 accuracy: 0.42 -> 53.15 | Improved |
| add channel-wise offset (ablation) | layer-wise IF baseline, ViT-L · ImageNet-1K T=64 | top-1 accuracy: 73.17 -> 84.49 | Improved |
| add channel-wise offset (ablation) | layer-wise IF baseline, ViT-L · ImageNet-1K T=8 | top-1 accuracy: 0.42 -> 66.34 | Improved |
| add channel-wise offset to channel-wise threshold (full AIF) (ablation) | CW-threshold, ViT-L · ImageNet-1K T=4 | top-1 accuracy: 1.73 -> 58.18 | Improved |
| add channel-wise threshold to channel-wise offset (full AIF) (ablation) | CW-offset, ViT-L · ImageNet-1K T=4 | top-1 accuracy: 0.17 -> 58.18 | Improved |
| add both components (full AIF) (ablation) | baseline, ViT-L · ImageNet-1K T=32 | top-1 accuracy: 63.45 -> 85.33 | Improved |
| increase time steps from 32 to 64 | AIF, ViT-B/16 · CIFAR10 / CIFAR100 | accuracy: 98.7 -> 98.9; 91.8 -> 92.0 | Improved |
| increase time steps from 8 to 64 | AIF, BERT-base · GLUE average | average score: 77.1 -> 82.7 | Improved |
| AIF conversion vs direct-training SpikeLM | AIF, BERT-base · GLUE average | average score: 76.5 -> 82.7 | Improved |
| AIF vs Spike-ZIP training-free conversion | AIF ViT-L · ImageNet-1K T=64 | top-1 accuracy: 83.82 -> 85.52 | Improved |
Closest papers by meaning
- Efficient ANN-SNN Conversion with Error Compensation Learning (ICML 2025 · similarity 0.92)
- Inference-Scale Complexity in ANN-SNN Conversion for High-Performance and Low-Power Applications (CVPR 2025 · similarity 0.92)
- Faster and Stronger: When ANN-SNN Conversion Meets Parallel Spiking Calculation (ICML 2025 · similarity 0.91)
- Adaptive Calibration: A Unified Conversion Framework of Spiking Neural Networks (AAAI 2025 · similarity 0.91)
- Temporal Misalignment in ANN-SNN Conversion and its Mitigation via Probabilistic Spiking Neurons (ICML 2025 · similarity 0.91)
- Differential Coding for Training-Free ANN-to-SNN Conversion (ICML 2025 · similarity 0.89)
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.
