Research map / Label-efficient and robust learning / Self-supervised and contrastive
Temperature Schedules for self-supervised contrastive methods on long-tail data
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 |
|---|---|---|---|
| Apply a cosine temperature schedule to MoCo | MoCo · CIFAR10-LT | FS LP: 69.31 -> 72.87 | Helped |
| Apply a cosine temperature schedule to MoCo | MoCo · CIFAR100-LT | FS LP: 26.86 -> 28.97 | Helped |
| Apply a cosine temperature schedule to SimCLR | SimCLR · CIFAR10-LT | kNN@1: 59.84 -> 63.09 | Helped |
| Apply a cosine temperature schedule to SimCLR | SimCLR · CIFAR100-LT | LT LP: 31.20 -> 33.28 | Helped |
| Apply a cosine temperature schedule to SimCLR | SimCLR · ImageNet100-LT | FS LP: 42.64 -> 45.18 | Helped |
| Apply a cosine temperature schedule rather than fixed temperature (ablation) | MoCo · CIFAR10-LT | FS LP: 68.89 -> 72.31 | Helped |
| Use a step schedule rather than fixed temperature (ablation) | MoCo · CIFAR10-LT | FS LP: 68.89 -> 70.18 | Helped |
| Sample temperature randomly rather than use fixed temperature (ablation) | MoCo · CIFAR10-LT | FS LP: 68.89 -> 70.26 | Helped |
| Use linear oscillation rather than fixed temperature (ablation) | MoCo · CIFAR10-LT | FS LP: 68.89 -> 71.50 | Helped |
| Use cosine schedule instead of linear oscillation (ablation) | MoCo · CIFAR10-LT | FS LP: 71.50 -> 72.31 | Helped |
| Widen the cosine schedule temperature range (ablation) | MoCo + TS · CIFAR10-LT | FS LP | Helped |
| Use cosine temperature scheduling on uniformly distributed rather than long-tailed data (ablation) | MoCo · CIFAR10-Uniform | FS LP: 90.19 -> 90.02 | No clear effect |
Closest papers by meaning
- Not All Semantics are Created Equal: Contrastive Self-supervised Learning with Automatic Temperature Individualization (ICML 2023 · similarity 0.89)
- Long-tailed Object Detection Pretraining: Dynamic Rebalancing Contrastive Learning with Dual Reconstruction (NeurIPS 2024 · similarity 0.87)
- Decoupled Contrastive Learning for Long-Tailed Recognition (AAAI 2024 · similarity 0.87)
- Continuous Contrastive Learning for Long-Tailed Semi-Supervised Recognition (NeurIPS 2024 · similarity 0.86)
- BaCon: Boosting Imbalanced Semi-supervised Learning via Balanced Feature-Level Contrastive Learning (AAAI 2024 · similarity 0.86)
- Long-Tailed Learning as Multi-Objective Optimization (AAAI 2024 · 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.