MetaCaDI: A Meta-Learning Framework for Causal Discovery from Multiple Environments with Unknown Interventions

Hans Jarett Ong, Yoichi Chikahara, Tomoharu Iwata
Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence, PMLR 337:5060-5080, 2026.

Abstract

Uncovering the causal mechanisms of complex real-world systems remains a significant challenge, as these systems often entail high data collection costs and involve unknown interventions. We introduce MetaCaDI, the first framework to cast the identification of unknown interventions as a meta-learning problem, explicitly leveraging a jointly learned causal graph. MetaCaDI is a {Bayesian} framework that learns a shared causal structure across multiple environments and is optimized to rapidly adapt to new, few-shot intervention target identification tasks. A key innovation is our model’s analytical adaptation, which uses a closed-form solution to bypass expensive and potentially unstable gradient-based bilevel optimization. Extensive experiments on synthetic and complex gene expression data demonstrate that MetaCaDI significantly outperforms state-of-the-art methods. It excels at identifying intervention targets from as few as 3 samples—where existing methods collapse to random chance—while robustly recovering the shared causal graph, proving its effectiveness in data-scarce scenarios.

Cite this Paper

BibTeX
@InProceedings{pmlr-v337-ong26a, title = {MetaCaDI: A Meta-Learning Framework for Causal Discovery from Multiple Environments with Unknown Interventions}, author = {Ong, Hans Jarett and Chikahara, Yoichi and Iwata, Tomoharu}, booktitle = {Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence}, pages = {5060--5080}, year = {2026}, editor = {Perković, Emilija and Malinsky, Daniel}, volume = {337}, series = {Proceedings of Machine Learning Research}, month = {17--21 Aug}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v337/main/assets/ong26a/ong26a.pdf}, url = {https://proceedings.mlr.press/v337/ong26a.html}, abstract = {Uncovering the causal mechanisms of complex real-world systems remains a significant challenge, as these systems often entail high data collection costs and involve unknown interventions. We introduce MetaCaDI, the first framework to cast the identification of unknown interventions as a meta-learning problem, explicitly leveraging a jointly learned causal graph. MetaCaDI is a {Bayesian} framework that learns a shared causal structure across multiple environments and is optimized to rapidly adapt to new, few-shot intervention target identification tasks. A key innovation is our model’s analytical adaptation, which uses a closed-form solution to bypass expensive and potentially unstable gradient-based bilevel optimization. Extensive experiments on synthetic and complex gene expression data demonstrate that MetaCaDI significantly outperforms state-of-the-art methods. It excels at identifying intervention targets from as few as 3 samples—where existing methods collapse to random chance—while robustly recovering the shared causal graph, proving its effectiveness in data-scarce scenarios.} }
Endnote
%0 Conference Paper %T MetaCaDI: A Meta-Learning Framework for Causal Discovery from Multiple Environments with Unknown Interventions %A Hans Jarett Ong %A Yoichi Chikahara %A Tomoharu Iwata %B Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2026 %E Emilija Perković %E Daniel Malinsky %F pmlr-v337-ong26a %I PMLR %P 5060--5080 %U https://proceedings.mlr.press/v337/ong26a.html %V 337 %X Uncovering the causal mechanisms of complex real-world systems remains a significant challenge, as these systems often entail high data collection costs and involve unknown interventions. We introduce MetaCaDI, the first framework to cast the identification of unknown interventions as a meta-learning problem, explicitly leveraging a jointly learned causal graph. MetaCaDI is a {Bayesian} framework that learns a shared causal structure across multiple environments and is optimized to rapidly adapt to new, few-shot intervention target identification tasks. A key innovation is our model’s analytical adaptation, which uses a closed-form solution to bypass expensive and potentially unstable gradient-based bilevel optimization. Extensive experiments on synthetic and complex gene expression data demonstrate that MetaCaDI significantly outperforms state-of-the-art methods. It excels at identifying intervention targets from as few as 3 samples—where existing methods collapse to random chance—while robustly recovering the shared causal graph, proving its effectiveness in data-scarce scenarios.
APA
Ong, H.J., Chikahara, Y. & Iwata, T.. (2026). MetaCaDI: A Meta-Learning Framework for Causal Discovery from Multiple Environments with Unknown Interventions. Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research 337:5060-5080 Available from https://proceedings.mlr.press/v337/ong26a.html.

Related Material