Contextual Causal Bayesian Optimisation
Organizations: CREST, ENSAE, IP Paris, France · Meta, FAIR · Huawei Noah’s Ark Lab, UK
Abstract
We introduce a unified framework for contextual and causal Bayesian optimisation, which aims to design intervention policies maximising the expectation of a target variable. Our approach leverages both observed contextual information and known causal graph structures to guide the search. Within this framework, we propose a novel algorithm that jointly optimises over policies and the sets of variables on which these policies are defined. This thereby extends and unifies two previously distinct approaches: Causal Bayesian Optimisation and Contextual Bayesian Optimisation, while also addressing their limitations in scenarios that yield suboptimal results. We derive worst-case and instance-dependent high-probability regret bounds for our algorithm. We report experimental results across diverse environments, corroborating that our approach achieves sublinear regret and reduces sample complexity in high-dimensional settings.
Figures & tables
| Acronym | Full text |
|---|---|
| BO | Bayesian Optimisation |
| CaBO | Causal Bayesian Optimisation |
| CoBO | Contextual Bayesian Optimisation |
| CoCa-BO | Contextual Causal Bayesian Optimisation |
| POMIS | Possibly Optimal Minimal Intervention Set |
| SCM | Structural Causal Model |
| Notation | Meaning |
|---|---|
| Vectors/sets composed of random variables | |
| Values taken by vectors/sets of random variables | |
| random variables | |
| The domains of the random variables ; i.e. , the sets in which take their values. | |
| The domains of the random vectors . | |
| A probability distribution of the random vector ; can be replaced by different symbols, for instance, by or by . |
Appendix figures & tables3 assets
Supplementary material from the paper’s appendix.
Appendix
| Kernel | , | |
|---|---|---|
| Matérn- 3 3 3 Please see Vakili et al. (2021) | ||
| Squared exponential 3 | ||
| Cauchy spectral mixture 4 4 4 Please see Zhang and Hua (2025) | ||
| -dimensional feature-map 5 5 5 Please see Srinivas et al. (2010) | ||
| -polynomial eigendecay 3 | ||
| -exponential eigendecay 3 |
Bayesian Optimization
Bayesian optimization uses probabilistic models and sequential experiments to find strong solutions when evaluating a black-box objective is costly.
See all 271 papers