Research map / Probabilistic and scientific modeling / Causal discovery
Interventions, Where and How? Experimental Design for Causal Models at Scale
Panagiotis Tigas, Yashas Annadani, Andrew Jesson, Bernhard Schölkopf, Yarin Gal, Stefan Bauer
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 |
|---|---|---|---|
| Select intervention values with GP-UCB instead of fixing the value or sampling from the observational support (ablation) | CBED · Synthetic 50-node Erdős–Rényi and Scale-Free graphs with nonlinear SCMs | E-SID | Improved |
| Sample intervention values from the observational support instead of using GP-UCB (ablation) | CBED · Synthetic 50-node Erdős–Rényi and Scale-Free graphs with nonlinear SCMs | E-SID | Hurt |
| Use soft top-k batch selection instead of greedy batch selection (ablation) | CBED · Synthetic 50-node Erdős–Rényi and Scale-Free graphs with nonlinear SCMs | E-SID | No clear effect |
| Select a batch of target-value pairs instead of repeating one selected pair for the whole batch (ablation) | CBED · Synthetic 50-node Erdős–Rényi and Scale-Free graphs with nonlinear SCMs | E-SID | Improved |
| Use Soft-CBED instead of random intervention acquisition | Causal discovery · DREAM Ecoli1 and Ecoli2 | E-SID | Improved |
| Use Soft-CBED instead of softAIT | Causal discovery · DREAM single-cell gene regulatory network datasets | E-SID | Improved |
| Use Soft-CBED with GP-UCB value selection instead of the evaluated baselines | Causal discovery · Synthetic 50-node Erdős–Rényi and Scale-Free graphs with nonlinear SCMs | E-SID | Improved |
| Use softAIT instead of AIT without soft batching (ablation) | AIT · Synthetic causal discovery tasks | E-SID | Improved |
| Use GP-UCB value selection rather than fixed values (ablation) | CBED · CBED experiments | runtime (s): 32.56 -> 284.98; 6.42 -> 24.17 | Hurt |
| Use soft batch selection instead of greedy batch selection (ablation) | CBED with fixed intervention values · CBED experiments | runtime (s): 32.56 -> 6.42 | Improved |
| Use soft batch selection instead of greedy batch selection (ablation) | CBED with GP-UCB value selection · CBED experiments | runtime (s): 284.98 -> 24.17 | Improved |
Closest papers by meaning
- Sample Efficient Bayesian Learning of Causal Graphs from Interventions (NeurIPS 2024 · similarity 0.90)
- Adaptive Online Experimental Design for Causal Discovery (ICML 2024 · similarity 0.90)
- When Selection Meets Intervention: Additional Complexities in Causal Discovery (ICLR 2025 · similarity 0.90)
- Challenges and Considerations in the Evaluation of Bayesian Causal Discovery (ICML 2024 · similarity 0.90)
- CausalTab: Pretraining Across Causal Environments for Tabular Causal Discovery (NeurIPS 2026 · similarity 0.90)
- Bivariate Causal Discovery using Bayesian Model Selection (ICML 2024 · similarity 0.90)
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.
