AI Paper Atlas

Research map / Label-efficient and robust learning / OOD and anomaly detection

Invariant Learning via Probability of Sufficient and Necessary Causes

Mengyue Yang, Zhen Fang, Yonggang Zhang, Yali Du, Furui Liu, Jean-Francois Ton, Jianhong Wang, Jun Wang

NeurIPS 2023 · Label-efficient and robust learning › OOD and anomaly detection · approach: Out-of-distribution generalization theory · 6 citations (OpenAlex)

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 changedWhereMeasuredEffect
add PNS-based CaSN risk to ERM-style training (CaSN base vs ERM)CaSN · PACSaverage accuracy: 85.5 -> 86.0Improved
add PNS risk (CaSN base vs ERM) on worst domainCaSN · PACSworst-domain accuracy: 79.3 -> 80.2Improved
add MMD loss to CaSN (ablation)CaSN · PACSworst-domain accuracy: 80.2 -> 81.4Improved
add IRM loss to CaSN (ablation)CaSN · PACSaverage accuracy: 86.0 -> 83.5Hurt
CaSN base vs ERM on VLCSCaSN · VLCSaverage accuracy: 77.5 -> 77.2No clear effect
add IRM loss to CaSN (ablation)CaSN · VLCSaverage accuracy: 77.2 -> 78.2Improved
add MMD loss to CaSN (ablation)CaSN · VLCSworst-domain accuracy: 64.8 -> 65.9Improved
CaSN vs ERM on ColoredMNISTCaSN · ColoredMNISTaverage accuracy: 51.5 -> 52.2Improved
remove Monotonicity evaluator (CaSN(-m)) (ablation)CaSN · synthetic datadistance correlation with SNHurt
larger delta (1.1 vs 0.1) (ablation)CaSN · synthetic datadistance correlation with SN: 0.90Improved
stronger spurious degree s=0.7 vs s=0.1CaSN · synthetic datadistance correlation with spurious factor: 0.13 -> 0.37Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy