Research map / Probabilistic and scientific modeling / Causal discovery
Causal Feature Learning via Generalized Rayleigh Quotients
Liang Cao, Jun Wan, Yan Qin, Weide Liu
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 lagged embedding (s=5 vs s=0) (ablation) | CFL · TEP (18 detectable faults) | average FDR: 92.5% -> 94.6% | Improved |
| add lagged embedding (s=5 vs s=0) (ablation) | CFL · TEP (18 detectable faults) | realized FAR: 2.1% -> 1.2% | Improved |
| increase lag order from 5 to 8 (ablation) | CFL · TEP (18 detectable faults) | average FDR: 94.6% -> 95.0% | No clear effect |
| increase stable dimension ds among 4, 8, 12 (ablation) | CFL · TEP (18 detectable faults) | average FDR: 94.5% to 94.6% | No clear effect |
| add varying monitor (dv=9 vs dv=0) (ablation) | CFL · TEP (18 detectable faults) | average FDR: 92.1% -> 94.6% | Improved |
| use dv=9 instead of dv=5 (ablation) | CFL · TEP (18 detectable faults) | average FDR: 92.9% -> 94.6% | Improved |
| use K=10 environments instead of K=3 (ablation) | CFL · TEP (18 detectable faults) | realized FAR: 3.2% -> 1.3% | Improved |
| use K=10 environments instead of K=20 (ablation) | CFL · TEP (18 detectable faults) | average FDR: 94.0% -> 94.6% | Improved |
| vary regularization epsilon from 1e-5 to 1e-2 (ablation) | CFL · TEP (18 detectable faults) | average FDR: 93.8% to 94.6% | No clear effect |
| add environment-mean invariance term Ae (ablation) | CFL · TEP (18 detectable faults) | average FDR: 93.6% -> 94.6% | Improved |
| add environment-mean invariance term Ae (ablation) | CFL · TEP (18 detectable faults) | realized FAR: 2.6% -> 1.3% | Improved |
| add target relevance term Ay (ablation) | CFL · TEP (18 detectable faults) | realized FAR: 2.6% -> 1.3% | Improved |
Closest papers by meaning
- Causality-Aware Contrastive Learning for Robust Multivariate Time-Series Anomaly Detection (ICML 2025 · similarity 0.85)
- Structured Temporal Causality for Interpretable Multivariate Time Series Anomaly Detection (NeurIPS 2025 · similarity 0.83)
- Task-Driven Causal Feature Distillation: Towards Trustworthy Risk Prediction (AAAI 2024 · similarity 0.83)
- GCAD: Anomaly Detection in Multivariate Time Series from the Perspective of Granger Causality (AAAI 2025 · similarity 0.83)
- Federated Granger Causality Learning For Interdependent Clients With State Space Representation (ICLR 2025 · similarity 0.82)
- A Sparsity Principle for Partially Observable Causal Representation Learning (ICML 2024 · similarity 0.82)
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.
