Estimating Causal Effects using a Multi-task Deep Ensemble

Ziyang Jiang, Zhuoran Hou, Yiling Liu, Yiman Ren, Keyu Li, David Carlson
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:15023-15040, 2023.

Abstract

A number of methods have been proposed for causal effect estimation, yet few have demonstrated efficacy in handling data with complex structures, such as images. To fill this gap, we propose Causal Multi-task Deep Ensemble (CMDE), a novel framework that learns both shared and group-specific information from the study population. We provide proofs demonstrating equivalency of CDME to a multi-task Gaussian process (GP) with a coregionalization kernel a priori. Compared to multi-task GP, CMDE efficiently handles high-dimensional and multi-modal covariates and provides pointwise uncertainty estimates of causal effects. We evaluate our method across various types of datasets and tasks and find that CMDE outperforms state-of-the-art methods on a majority of these tasks.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-jiang23c, title = {Estimating Causal Effects using a Multi-task Deep Ensemble}, author = {Jiang, Ziyang and Hou, Zhuoran and Liu, Yiling and Ren, Yiman and Li, Keyu and Carlson, David}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {15023--15040}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/jiang23c/jiang23c.pdf}, url = {https://proceedings.mlr.press/v202/jiang23c.html}, abstract = {A number of methods have been proposed for causal effect estimation, yet few have demonstrated efficacy in handling data with complex structures, such as images. To fill this gap, we propose Causal Multi-task Deep Ensemble (CMDE), a novel framework that learns both shared and group-specific information from the study population. We provide proofs demonstrating equivalency of CDME to a multi-task Gaussian process (GP) with a coregionalization kernel a priori. Compared to multi-task GP, CMDE efficiently handles high-dimensional and multi-modal covariates and provides pointwise uncertainty estimates of causal effects. We evaluate our method across various types of datasets and tasks and find that CMDE outperforms state-of-the-art methods on a majority of these tasks.} }
Endnote
%0 Conference Paper %T Estimating Causal Effects using a Multi-task Deep Ensemble %A Ziyang Jiang %A Zhuoran Hou %A Yiling Liu %A Yiman Ren %A Keyu Li %A David Carlson %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-jiang23c %I PMLR %P 15023--15040 %U https://proceedings.mlr.press/v202/jiang23c.html %V 202 %X A number of methods have been proposed for causal effect estimation, yet few have demonstrated efficacy in handling data with complex structures, such as images. To fill this gap, we propose Causal Multi-task Deep Ensemble (CMDE), a novel framework that learns both shared and group-specific information from the study population. We provide proofs demonstrating equivalency of CDME to a multi-task Gaussian process (GP) with a coregionalization kernel a priori. Compared to multi-task GP, CMDE efficiently handles high-dimensional and multi-modal covariates and provides pointwise uncertainty estimates of causal effects. We evaluate our method across various types of datasets and tasks and find that CMDE outperforms state-of-the-art methods on a majority of these tasks.
APA
Jiang, Z., Hou, Z., Liu, Y., Ren, Y., Li, K. & Carlson, D.. (2023). Estimating Causal Effects using a Multi-task Deep Ensemble. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:15023-15040 Available from https://proceedings.mlr.press/v202/jiang23c.html.

Related Material