Research map / Probabilistic and scientific modeling / Causal discovery
Learning causal graphs via nonlinear sufficient dimension reduction
Eftychia Solea, Bing Li, Kyongwon Kim
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 nonlinear sufficient dimension reduction before the kernel conditional independence test (DAG-PC vs Kernel-PC) (ablation) | Kernel-PC · Gaussian DAGs, n=200 p=50, E[Ni]=2 | CPDAG AUC: 0.73(0.02) -> 0.81(0.03) | Improved |
| add nonlinear sufficient dimension reduction (DAG-PC vs Kernel-PC) (ablation) | Kernel-PC · Non-Gaussian DAGs, n=200 p=50, E[Ni]=2 | CPDAG AUC: 0.61(0.02) -> 0.63(0.02) | Improved |
| add nonlinear sufficient dimension reduction (DAG-PC vs Kernel-PC) (ablation) | Kernel-PC · Mixed non-Gaussian DAGs, n=200 p=50, E[Ni]=2 | CPDAG AUC: 0.61(0.03) -> 0.64(0.01) | Improved |
| use DAG-PC instead of Gaussian-PC on nonlinear data | Gaussian-PC · Non-Gaussian DAGs, n=200 p=50, E[Ni]=4 | skeleton AUC: 0.58(0.02) -> 0.68(0.02) | Improved |
| use DAG-PC instead of Gaussian-PC on Gaussian data | Gaussian-PC · Gaussian DAGs, n=200 p=50, E[Ni]=2 | skeleton AUC: 0.90(0.02) -> 0.87(0.03) | Hurt |
| use DAG-PC instead of linear PC methods in denser Gaussian graphs (E[Ni]=4) | Gaussian-PC and Rank-PC · Gaussian DAGs, E[Ni]=4 | SHD | Hurt |
| increase network size to p=50 (ablation) | Gaussian-PC, Rank-PC, Kernel-PC · Non-Gaussian and mixed DAGs | SHD / AUC | Hurt |
| use dimension of sufficient predictor 2 instead of 1 (ablation) | DAG-PC · Simulated DAGs | SHD / AUC | No clear effect |
| increase sample size from 100 to 200 for DAG-PC (ablation) | DAG-PC · Non-Gaussian DAGs, p=5, E[Ni]=2 | CPDAG AUC: 0.62(0.08) -> 0.70(0.11) | Improved |
Closest papers by meaning
- Coarsening Linear Non-Gaussian Causal Models with Cycles (NeurIPS 2026 · similarity 0.91)
- Effective Causal Discovery under Identifiable Heteroscedastic Noise Model (AAAI 2024 · similarity 0.89)
- ProDAG: Projected Variational Inference for Directed Acyclic Graphs (NeurIPS 2025 · similarity 0.89)
- Sample Efficient Bayesian Learning of Causal Graphs from Interventions (NeurIPS 2024 · similarity 0.89)
- Causal Discovery with Fewer Conditional Independence Tests (ICML 2024 · similarity 0.89)
- A Fixed-Parameter Tractable Algorithm for Counting Markov Equivalence Classes with the Same Skeleton (AAAI 2024 · similarity 0.89)
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.
