Research map / Video / Video segmentation and tracking
PredNext: Explicit Cross-View Temporal Prediction for Unsupervised Learning in Spiking Neural Networks
Yiting Dong, Jianhao Ding, Zijie Xu, Tong Bu, Zhaofei Yu, Tiejun Huang
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 PredNext to SimCLR | SimCLR SNN · UCF101 (pretrain UCF101) | top-1 accuracy: 57.04 -> 59.47 | Improved |
| add PredNext to SimSiam | SimSiam SNN · UCF101 (pretrain miniKinetics) | top-1 accuracy: 43.77 -> 50.65 | Improved |
| add PredNext to MoCo | MoCo SNN · UCF101 (pretrain UCF101) | top-1 accuracy: 49.70 -> 54.98 | Improved |
| add PredNext to SimSiam | SimSiam SNN · HMDB51 (pretrain miniKinetics) | top-1 accuracy: 19.08 -> 25.03 | Improved |
| add PredNext to SimCLR | SimCLR SNN · HMDB51 (pretrain miniKinetics) | top-1 accuracy: 35.42 -> 36.80 | No clear effect |
| force temporal consistency directly with a loss (beta=0.8) instead of PredNext (ablation) | SimSiam (ImageNet) · UCF101 | top-1 accuracy: 70.32 -> 60.35 | Hurt |
| add clip prediction (ablation) | PredNext (SimSiam) · UCF101 | top-1 accuracy: 50.81 -> 54.40 | Improved |
| add step prediction (ablation) | SimSiam · UCF101 | top-1 accuracy: 50.81 -> 51.33 | Improved |
| use cross-view instead of same-view prediction (ablation) | PredNext (SimSiam) · UCF101 | top-1 accuracy: 53.66 -> 54.93 | Improved |
| remove original SSL loss, same-view prediction only (ablation) | PredNext (SimSiam) · UCF101 | top-1 accuracy: 54.93 -> 5.03 | Hurt |
| raise prediction loss weight alpha from 0.5 to 1.0 (ablation) | PredNext (SimCLR) · UCF101 | top-1 accuracy: 59.5 -> 52.4 | Hurt |
| add PredNext to SimSiam (video retrieval) | SimSiam · UCF101 retrieval | R@1: 27.84 -> 36.27 | Improved |
Closest papers by meaning
- Temporal Flexibility in Spiking Neural Networks: Towards Generalization Across Time Steps and Deployment Friendliness (ICLR 2025 · similarity 0.87)
- Towards More Discriminative Feature Learning in SNNs with Temporal-Self-Erasing Supervision (AAAI 2025 · similarity 0.87)
- Learning Delays in Spiking Neural Networks using Dilated Convolutions with Learnable Spacings (ICLR 2024 · similarity 0.86)
- Temporal Separation with Entropy Regularization for Knowledge Distillation in Spiking Neural Networks (CVPR 2025 · similarity 0.86)
- STAA-SNN: Spatial-Temporal Attention Aggregator for Spiking Neural Networks (CVPR 2025 · similarity 0.86)
- SpikingSSMs: Learning Long Sequences with Sparse and Parallel Spiking State Space Models (AAAI 2025 · similarity 0.86)
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.
