Research map / Architectures and efficiency / Spiking networks
Error-Aware Conversion from ANN to SNN via Post-training Parameter Calibration
Yuhang Li, Shikuang Deng, Xin Dong, Shi Gu
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 ClipRound (bias offset) instead of floor (ablation) | MaxAct SNN, VGG-16 no BN, T=32 · ImageNet | top-1 accuracy: 0.1 -> 63.22 | Improved |
| add ClipRound instead of floor (ablation) | MMSE SNN, ResNet-34 no BN, T=16 · ImageNet | top-1 accuracy: 5.54 -> 28.70 | Improved |
| use MMSE threshold instead of percentile (ablation) | SNN ResNet-34 no BN, T=16 · ImageNet | top-1 accuracy: 15.09 -> 28.70 | Improved |
| use MMSE threshold instead of maximum activation (ablation) | SNN ResNet-34 no BN, T=16 · ImageNet | top-1 accuracy: 7.46 -> 28.70 | Improved |
| use channel-wise MMSE threshold instead of layer-wise (ablation) | SNN ResNet-34 no BN, T=16 · ImageNet | top-1 accuracy: 28.70 -> 43.69 | Improved |
| add light calibration pipeline (bias calibration) (ablation) | baseline SNN VGG-16 with BN, T=16 · ImageNet | top-1 accuracy: 0.95 -> 40.81 | Improved |
| add potential and weight calibration (ACP) on top of LCP (ablation) | SNN VGG-16 with BN, T=16 · ImageNet | top-1 accuracy: 40.81 -> 65.02 | Improved |
| combine ACP with maximum activation threshold instead of MMSE (ablation) | SNN ResNet-34 with BN, T=16 · ImageNet | top-1 accuracy: 50.84 -> 0.41 | Hurt |
| use ANN with BN layers instead of without BN, no calibration (ablation) | MMSE + ClipRound SNN ResNet-34, T=16 · ImageNet | top-1 accuracy: 28.70 -> 0.10 | Hurt |
| add ACP calibration (ablation) | SNN MobileNet, T=128 · ImageNet | top-1 accuracy: 1.99 -> 64.58 | Improved |
| add ACP calibration (ablation) | SNN RetinaNet, T=32 · COCO | mAP: 6.2 -> 28.4 | Improved |
| increase LCP calibration samples (ablation) | LCP ResNet-34 · ImageNet | top-1 accuracy: 57.23 -> 66.23 | Improved |
Closest papers by meaning
- Efficient ANN-SNN Conversion with Error Compensation Learning (ICML 2025 · similarity 0.93)
- Adaptive Calibration: A Unified Conversion Framework of Spiking Neural Networks (AAAI 2025 · similarity 0.92)
- Inference-Scale Complexity in ANN-SNN Conversion for High-Performance and Low-Power Applications (CVPR 2025 · similarity 0.91)
- Temporal Misalignment in ANN-SNN Conversion and its Mitigation via Probabilistic Spiking Neurons (ICML 2025 · similarity 0.90)
- Faster and Stronger: When ANN-SNN Conversion Meets Parallel Spiking Calculation (ICML 2025 · similarity 0.90)
- Self-Attentive Spatio-Temporal Calibration for Precise Intermediate Layer Matching in ANN-to-SNN Distillation (AAAI 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.
