Research map / Probabilistic and scientific modeling / Causal discovery
Integer Programming for Causal Structure Learning in the Presence of Latent Variables
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 |
|---|---|---|---|
| Allow c-components beyond single-node districts through the AGIP formulation | ADMG structure learning · Random ADMGs overall | SHD: 25.5 (AGIP) vs 28.8 (M³HC); 25.5 (AGIP) vs 31.2 (FCI) | Helped |
| Use exact score-based AGIP rather than greedy M³HC | ADMG structure learning · Randomly generated instances | solution score: AGIP strictly better on 53 instances; M³HC slightly better on 4 of 60 | Helped |
| Allow non-DAG c-components in AGIP rather than restricting to DAGIP | ADMG structure learning · Non-DAG-representable ADMGs overall | precision: 46.3 -> 49.6 | Helped |
| Allow non-DAG c-components in AGIP rather than restricting to DAGIP | ADMG structure learning · Non-DAG-representable ADMGs overall | recall: 43.1 -> 47.2 | Helped |
| Increase the number of latent variables from 2 to 6 | AGIP · d=18, N=10000 random ADMGs | SHD: 12.6 -> 28.8 | Hurt |
| Use AGIP rather than DAGIP | ADMG structure learning · Random ADMGs, 60 instances | solution score: AGIP improves on 13 of 60 instances; scores tie on the remaining instances | Mixed |
| Add candidate c-components implied by three-node districts, with up to 1 parent per node | AGIP · Non-DAG-representable ADMGs | solution score: AGIP strictly better than DAGIP on 28 out of 50 instances | Helped |
| Restrict the candidate c-component set to the district and parent sizes described in the experiments | AGIP · Random ADMGs as latent-variable count increases | SHD | Hurt |
| Increase the number of latent variables from 2 to 6 | AGIP · d=18, N=10000 random ADMGs | precision: 80.3 -> 57.8 | Hurt |
Closest papers by meaning
- Differentiable Structure Learning with Ancestral Constraints (ICML 2025 · similarity 0.90)
- Boosting Causal Structure Learning: An Asymmetric Exponential Modulation Gaussian-Based Adaptive Sample Reweighting Framework (AAAI 2025 · similarity 0.89)
- Effective Causal Discovery under Identifiable Heteroscedastic Noise Model (AAAI 2024 · similarity 0.89)
- Scores for Learning Discrete Causal Graphs with Unobserved Confounders (AAAI 2024 · similarity 0.88)
- Causal Discovery via Bayesian Optimization (ICLR 2025 · similarity 0.88)
- Interventional Causal Discovery in a Mixture of DAGs (NeurIPS 2024 · similarity 0.88)
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.