Research map / Label-efficient and robust learning / OOD and anomaly detection
The Missing Invariance Principle found -- the Reciprocal Twin of Invariant Risk Minimization
Dongsung Huh, Avinash Baidya
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 |
|---|---|---|---|
| replace IRM-v1 with MRI-v1 (PM) | invariance penalty · S-T Regression test | risk: 1.00 -> 0.54 | Improved |
| replace IRM-v1 with MRI-v1 (ALM) | invariance penalty · S-T Regression test | risk: 0.62 -> 0.46 | Improved |
| replace IRM-v1 with MRI-v1 (ALM) | invariance penalty · S-T Classification test | accuracy: 0.77 -> 0.86 | Improved |
| replace IRM-v1 with MRI-v1 (PM) | invariance penalty · S-T Classification test | accuracy: 0.52 -> 0.85 | Improved |
| replace IRM-v1 with MRI-v1 (ALM) | invariance penalty · CMNISTa test | risk: 0.69 -> 0.64 | Improved |
| replace IRM-v1 with MRI-v1 (ALM) | invariance penalty · CMNISTb test | risk: 0.47 -> 0.29 | Improved |
| replace IRM-v1 with MRI-v1 (ALM) | invariance penalty · CMNISTa test | accuracy: 0.66 -> 0.65 | No clear effect |
| replace IRM-v1 with MRI-v1 (ALM) | invariance penalty · CMNISTb test | accuracy: 0.91 -> 0.9 | No clear effect |
| add invariance constraint (IRM-v1 PM) over ERM | ERM · CMNISTa test | accuracy: 0.10 -> 0.23 | Mixed |
| use MRI-v1 instead of ERM | ERM · S-T Classification test | accuracy: 0.50 -> 0.86 | Improved |
| switch penalty method (PM) to augmented Lagrangian (ALM) (ablation) | IRM-v1 · S-T Classification test | accuracy: 0.52 -> 0.77 | Improved |
| switch penalty method (PM) to augmented Lagrangian (ALM) (ablation) | MRI-v1 · CMNISTb test | risk: 0.4 -> 0.29 | Improved |
Closest papers by meaning
- Learning Optimal Features via Partial Invariance (AAAI 2023 · similarity 0.91)
- The Risks of Invariant Risk Minimization (ICLR 2021 · similarity 0.91)
- Invariant Risk Minimization Is A Total Variation Model (ICML 2024 · similarity 0.88)
- Diagnosing and Rectifying Fake OOD Invariance: A Restructured Causal Approach (AAAI 2024 · similarity 0.87)
- The Implicit Bias of Heterogeneity towards Invariance: A Study of Multi-Environment Matrix Sensing (NeurIPS 2024 · similarity 0.85)
- Sufficient Invariant Learning for Distribution Shift (CVPR 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.
