Research map / Probabilistic and scientific modeling / Bayesian optimization
Constrained Causal Bayesian Optimization
Virginia Aglietti, Alan Malek, Ira Ktena, Silvia Chiappa
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 observational data to construct GP prior parameters | cCBO · Synthetic-1 and Synthetic-2, N_O=500 | convergence speed | Improved |
| Capture covariance between target and constraint effects | cCBO with observational-data priors · Synthetic-1, N_O=500 | convergence speed | Improved |
| Use the SCM structure to model effect covariance | cCBO with G-mtgp · Synthetic-2, N_O=500 | feasible interventions collected | Improved |
| Use observational-data priors when the observational sample is small | cCBO with G-mtgp · Health, N_O=10 | convergence | Hurt |
| Use observational-data priors when the observational sample is small | cCBO with mtgp+ · Synthetic-2, N_O=100 and N_O=10 | convergence speed | Hurt |
| Use mtgp+ rather than G-mtgp when the observational SCM fit is inaccurate | cCBO · Synthetic-1, N_O=100 and N_O=10 | convergence speed | Improved |
| Use mtgp+ rather than observational-data-based priors with a very small observational sample | cCBO · Synthetic-2, N_O=100 and N_O=10 | convergence speed | Improved |
| Use G-mtgp rather than cBO | constrained optimization · Health, N_O=100 | convergence speed | Improved |
| Use G-mtgp rather than cBO | constrained optimization · Health, N_O=100 | feasible interventions collected | Improved |
| Intervene on all variables instead of selecting causally relevant intervention sets | cBO · Synthetic-1 | target effect | Hurt |
| Model target and constraint effects independently rather than capturing their correlation (ablation) | cCBO with stgp+ · Synthetic-2, N_O=500 | feasible interventions collected | Hurt |
| Use observational-data priors rather than zero-mean GP priors (ablation) | cCBO · Synthetic-1 and Synthetic-2, N_O=500 | convergence speed | Improved |
Closest papers by meaning
- Adversarial Causal Bayesian Optimization (ICLR 2024 · similarity 0.89)
- Sample Efficient Bayesian Learning of Causal Graphs from Interventions (NeurIPS 2024 · similarity 0.89)
- Causal Discovery via Bayesian Optimization (ICLR 2025 · similarity 0.88)
- A General Framework for User-Guided Bayesian Optimization (ICLR 2024 · similarity 0.87)
- Intervention and Conditioning in Causal Bayesian Networks (NeurIPS 2024 · similarity 0.87)
- Bayesian Active Learning for Bivariate Causal Discovery (ICML 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.
