Research map / Label-efficient and robust learning / Self-supervised and contrastive
Rethinking Generalizability and Discriminability of Self-Supervised Learning from Evolutionary Game Theory Perspective
Jiangmeng Li, Zehua Zang, Qirui Ji, Chuxiong Sun, Wenwen Qiang, Junge Zhang, Changwen Zheng, Fuchun Sun, Hui Xiong
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 EGT-derived loss weights instead of plain sum of losses (ablation) | ESSL · Tiny-ImageNet (pre-trained on STL10) | top-1 accuracy: 27.5 -> 29.7 | Improved |
| add EGT-derived loss weights instead of plain sum of losses (ablation) | ESSL · CIFAR100 (pre-trained on CIFAR100) | top-1 accuracy: 47.1 -> 49.9 | Improved |
| add EGT-derived loss weights instead of plain sum of losses (ablation) | ESSL · CIFAR10 (pre-trained on CIFAR10) | top-1 accuracy: 79.0 -> 83.5 | Improved |
| add RL-adapted weights on top of static EGT weights (ablation) | ESSL · CIFAR100 (pre-trained on STL10) | top-1 accuracy: 41.4 -> 45.4 | Improved |
| add RL-adapted weights on top of static EGT weights (ablation) | ESSL · Tiny-ImageNet (pre-trained on Tiny-ImageNet) | top-1 accuracy: 35.1 -> 32.0 | Hurt |
| add RL-adapted weights on top of static EGT weights (ablation) | ESSL · CIFAR10 (pre-trained on CIFAR10) | top-1 accuracy: 83.5 -> 84.3 | Improved |
| use two projection heads instead of one (ablation) | ESSL · CIFAR10 (pre-trained on CIFAR10) | top-1 accuracy: 83.5 -> 67.3 | Hurt |
| use two projection heads instead of one (ablation) | ESSL · CIFAR100 (pre-trained on CIFAR100) | top-1 accuracy: 49.9 -> 35.9 | Hurt |
| plain sum of losses versus Barlow Twins alone (ablation) | ESSL w/ Ensemble · STL10 (pre-trained on STL10) | top-1 accuracy: 85.5 -> 84.0 | Hurt |
| add Barlow Twins loss to DINO (ablation) | ESSL(DINO+Barlow Twins) · STL10 (pre-trained on STL10) | top-1 accuracy: 85.8 -> 88.4 | Improved |
| add Barlow Twins loss to SimCLR-dagger (ablation) | ESSL(SimCLR+Barlow Twins) · CIFAR10 (pre-trained on CIFAR10) | top-1 accuracy: 74.4 -> 84.3 | Improved |
| add Barlow Twins loss to Barlow Twins baseline pairing with SimCLR-dagger (ablation) | ESSL(SimCLR+Barlow Twins) · CIFAR100 (pre-trained on CIFAR100) | top-1 accuracy: 51.1 -> 51.1 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Barlow Twins — Barlow Twins: Self-Supervised Learning via Redundancy Reduction (ICML 2021) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Towards Generalizable Reinforcement Learning via Causality-Guided Self-Adaptive Representations (ICLR 2025 · similarity 0.87)
- Generalizing Reinforcement Learning through Fusing Self-Supervised Learning into Intrinsic Motivation (AAAI 2022 · similarity 0.87)
- Learning Diverse Risk Preferences in Population-Based Self-Play (AAAI 2024 · similarity 0.85)
- On Generalization Across Environments In Multi-Objective Reinforcement Learning (ICLR 2025 · similarity 0.85)
- Bridging State and History Representations: Understanding Self-Predictive RL (ICLR 2024 · similarity 0.84)
- An Optimal Discriminator Weighted Imitation Perspective for Reinforcement Learning (ICLR 2025 · similarity 0.84)
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.
