Research map / Probabilistic and scientific modeling / Causal discovery
Dissecting Causal Mechanism Shifts via FANS: Function And Noise Separation
Gyeongdeok Seo, Jaeyoon Shim, Mingyu Kim, Hoyoon Byun, Yonghan Jung, Kyungwoo Song
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 |
|---|---|---|---|
| use FANS instead of PreDITEr | synthetic SCMs · Shifted node detection ER J=10 | F1: 0.940 -> 0.989 | Improved |
| use FANS instead of PreDITEr | synthetic SCMs · Shifted node detection SF J=10 | F1: 0.933 -> 1.000 | Improved |
| use FANS instead of GPR for shift type dissection | synthetic SCMs · Dissection SF J=10 | F1: 0.722 -> 1.000 | Improved |
| use FANS instead of GPR | synthetic SCMs · Dissection ER J=50 | F1: 0.906 -> 0.947 | Improved |
| use FANS instead of GPR for function-only vs simultaneous shifts | GAM-based noise standardization · simultaneous shift | macro F1: 0.378 -> 0.622 | Improved |
| use FANS instead of iSCAN | Morpho-MNIST · top-100 shifted pixels | pixels detected: 3 -> 57 | Improved |
| use elbow rule instead of fixed threshold tau=0.1 (ablation) | FANS detection · SF graph J=50 | F1: 1.000 -> 0.926 | Hurt |
| use elbow rule instead of fixed threshold tau=0.1 (ablation) | FANS detection · ER graph J=30 | F1: 0.980 -> 0.992 | No clear effect |
| reduce CNF hidden dimensions from [32,32,32] to [8,8] (ablation) | FANS · synthetic SCMs | detection and dissection F1 | Hurt |
| vary batch size and learning rate (ablation) | FANS · synthetic SCMs | F1 | No clear effect |
| reduce training sample size down to N=5000 (ablation) | FANS · synthetic SCMs | F1 | No clear effect |
Closest papers by meaning
- Identifying Structural Biases from Causal Mechanism Shifts (NeurIPS 2026 · similarity 0.88)
- Causal Discovery via Transformed Low-Rank Quantile Surfaces (NeurIPS 2026 · similarity 0.86)
- A Skewness-Based Criterion for Addressing Heteroscedastic Noise in Causal Discovery (ICLR 2025 · similarity 0.86)
- CausalDynamics: A large‐scale benchmark for structural discovery of dynamical causal models (NeurIPS 2025 · similarity 0.86)
- Distinguishing Cause from Effect with Causal Velocity Models (ICML 2025 · similarity 0.86)
- Identifying Macro Causal Effects in C-DMGs over DMGs (NeurIPS 2025 · similarity 0.86)
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.
