Research map / Probabilistic and scientific modeling / Causal discovery
SELF: Structural Equational Likelihood Framework for Causal Discovery
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 a parent node versus a non-parent node to a correct parent set | SELF noise-entropy gap (Assumption 1) · Synthetic linear and nonlinear data, varied signal-to-noise ratio | entropy gap | Helped |
| increase the number of Markov equivalence edges in the random structures | SELF, LiNGAM, DLiNGAM · Synthetic linear data, 4000 samples, avg indegree 1.5 | F1 | No clear effect |
| increase the number of Markov equivalence edges in the random structures | HCBN (traditional likelihood-based hill climbing) · Synthetic linear data, 4000 samples, avg indegree 1.5 | F1 | Hurt |
| increase the number of Markov equivalence edges in the random structures | SELF versus MMPC-ANM · Synthetic nonlinear data, 4000 samples, avg indegree 1.5 | F1 | No clear effect |
| increase the sample size from 1000 up to 8000 | SELF and linear baselines · Synthetic linear data, 40 variables, avg indegree 1.5 | F1 | No clear effect |
| increase the number of variables from 20 to 60 | SELF versus LiNGAM/DLiNGAM · Synthetic linear data, 4000 samples, avg indegree 1.5 | F1 | Helped |
| increase the average indegree from 0.5 to 2.5 | SELF and HCBN versus LiNGAM and DLiNGAM · Synthetic linear data, 40 variables, 4000 samples | F1 | Mixed |
| increase the sample size from 2000 to 8000 | SELF (XGBoost-GBTree) · Synthetic nonlinear data, 40 variables, avg indegree 1.5 | F1 | Helped |
| use SELF instead of MMPC-ANM on nonlinear data across variables and indegree | Structure learning method · Synthetic nonlinear data | F1 | Helped |
| use SELF instead of LiNGAM, DLiNGAM and HCBN on real-world structures | Structure learning method · Child, Alarm, Win95pts, Pathfinder (linear and nonlinear data) | recall, precision, F1 | Helped |
| increase the scale of the real-world structure | SELF advantage over baselines · Real-world structures | F1 | Helped |
| compare precision gain with recall gain of SELF over baselines | SELF · Real-world structures, linear data | precision vs recall | Helped |
Closest papers by meaning
- Bivariate Causal Discovery using Bayesian Model Selection (ICML 2024 · similarity 0.89)
- Hybrid Top-Down Global Causal Discovery with Local Search for Linear and Nonlinear Additive Noise Models (NeurIPS 2024 · similarity 0.88)
- Strong and Weak Identifiability of Optimization-based Causal Discovery in Non-linear Additive Noise Models (ICML 2025 · similarity 0.88)
- Local Causal Structure Learning in the Presence of Latent Variables (ICML 2024 · similarity 0.87)
- A Meta-Learning Approach to Bayesian Causal Discovery (ICLR 2025 · similarity 0.87)
- Differentiable Causal Discovery for Latent Hierarchical Causal Models (ICLR 2025 · similarity 0.87)
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.