Research map / Label-efficient and robust learning / Self-supervised and contrastive
Self-Supervised Learning Disentangled Group Representation as Feature
Tan Wang, Zhongqi Yue, Jianqiang Huang, Qianru Sun, Hanwang Zhang
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 IP-IRM to HCL | HCL · STL10 400 epochs | linear accuracy: 87.60 -> 87.81 | Improved |
| add IP-IRM to HCL | HCL · STL10 400 epochs | k-NN accuracy: 80.06 -> 84.29 | Improved |
| add IP-IRM to SimCLR | SimCLR · STL10 400 epochs | linear accuracy: 78.89 -> 84.44 | Improved |
| add IP-IRM to SimCLR | SimCLR · Cifar100 400 epochs | linear accuracy: 66.63 -> 69.55 | Improved |
| add IP-IRM to DCL | DCL · Cifar100 400 epochs | linear accuracy: 68.59 -> 68.76 | Improved |
| add IP-IRM to SimCLR with MixUp, 1000 epochs | SimCLR · STL10 | k-NN accuracy: 79.80 -> 85.08 | Improved |
| add IP-IRM to SimCLR with MixUp, 1000 epochs | SimCLR · Cifar100 | linear accuracy: 72.18 -> 73.99 | Improved |
| add IP-IRM to MoCo-v2 | MoCo-v2 · few-shot 5-way-5-shot average | accuracy: 67.22 -> 69.44 | Improved |
| IP-IRM vs SSL baselines on NICO | SSL baselines · NICO | accuracy: 1.5-4.2% | Improved |
| IP-IRM vs SimCLR on disentanglement | SimCLR · Shapes3D | average score: 0.645 -> 0.672 | Improved |
| IP-IRM vs SimCLR on disentanglement | SimCLR · CMNIST | average score: 0.873 -> 0.906 | Improved |
| set lambda1 and lambda2 to 1.0 instead of 0.2-0.5 (ablation) | IP-IRM · STL10 and Cifar100, 200 epochs | linear accuracy | Hurt |
Closest papers by meaning
- Towards Building A Group-based Unsupervised Representation Disentanglement Framework (ICLR 2022 · similarity 0.87)
- Learning Disentangled Representation with Pairwise Independence (AAAI 2019 · similarity 0.86)
- Disentangled Representation Learning via Modular Compositional Bias (NeurIPS 2025 · similarity 0.86)
- Transferring disentangled representations: bridging the gap between synthetic and real images (NeurIPS 2024 · similarity 0.85)
- An Information Criterion for Controlled Disentanglement of Multimodal Data (ICLR 2025 · similarity 0.85)
- Disentangled Representation Learning with the Gromov-Monge Gap (ICLR 2025 · similarity 0.85)
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.
