Amortized Causal Discovery: Learning to Infer Causal Graphs from Time-Series Data

Sindy Löwe, David Madras, Richard Zemel, Max Welling
Proceedings of the First Conference on Causal Learning and Reasoning, PMLR 177:509-525, 2022.

Abstract

On time-series data, most causal discovery methods fit a new model whenever they encounter samples from a new underlying causal graph. However, these samples often share relevant information which is lost when following this approach. Specifically, different samples may share the dynamics which describe the effects of their causal relations. We propose Amortized Causal Discovery, a novel framework that leverages such shared dynamics to learn to infer causal relations from time-series data. This enables us to train a single, amortized model that infers causal relations across samples with different underlying causal graphs, and thus leverages the shared dynamics information. We demonstrate experimentally that this approach, implemented as a variational model, leads to significant improvements in causal discovery performance, and show how it can be extended to perform well under added noise and hidden confounding.

Cite this Paper

BibTeX
@InProceedings{pmlr-v177-lowe22a, title = {Amortized Causal Discovery: Learning to Infer Causal Graphs from Time-Series Data}, author = {L{\"o}we, Sindy and Madras, David and Zemel, Richard and Welling, Max}, booktitle = {Proceedings of the First Conference on Causal Learning and Reasoning}, pages = {509--525}, year = {2022}, editor = {Schölkopf, Bernhard and Uhler, Caroline and Zhang, Kun}, volume = {177}, series = {Proceedings of Machine Learning Research}, month = {11--13 Apr}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v177/lowe22a/lowe22a.pdf}, url = {https://proceedings.mlr.press/v177/lowe22a.html}, abstract = {On time-series data, most causal discovery methods fit a new model whenever they encounter samples from a new underlying causal graph. However, these samples often share relevant information which is lost when following this approach. Specifically, different samples may share the dynamics which describe the effects of their causal relations. We propose Amortized Causal Discovery, a novel framework that leverages such shared dynamics to learn to infer causal relations from time-series data. This enables us to train a single, amortized model that infers causal relations across samples with different underlying causal graphs, and thus leverages the shared dynamics information. We demonstrate experimentally that this approach, implemented as a variational model, leads to significant improvements in causal discovery performance, and show how it can be extended to perform well under added noise and hidden confounding.} }
Endnote
%0 Conference Paper %T Amortized Causal Discovery: Learning to Infer Causal Graphs from Time-Series Data %A Sindy Löwe %A David Madras %A Richard Zemel %A Max Welling %B Proceedings of the First Conference on Causal Learning and Reasoning %C Proceedings of Machine Learning Research %D 2022 %E Bernhard Schölkopf %E Caroline Uhler %E Kun Zhang %F pmlr-v177-lowe22a %I PMLR %P 509--525 %U https://proceedings.mlr.press/v177/lowe22a.html %V 177 %X On time-series data, most causal discovery methods fit a new model whenever they encounter samples from a new underlying causal graph. However, these samples often share relevant information which is lost when following this approach. Specifically, different samples may share the dynamics which describe the effects of their causal relations. We propose Amortized Causal Discovery, a novel framework that leverages such shared dynamics to learn to infer causal relations from time-series data. This enables us to train a single, amortized model that infers causal relations across samples with different underlying causal graphs, and thus leverages the shared dynamics information. We demonstrate experimentally that this approach, implemented as a variational model, leads to significant improvements in causal discovery performance, and show how it can be extended to perform well under added noise and hidden confounding.
APA
Löwe, S., Madras, D., Zemel, R. & Welling, M.. (2022). Amortized Causal Discovery: Learning to Infer Causal Graphs from Time-Series Data. Proceedings of the First Conference on Causal Learning and Reasoning, in Proceedings of Machine Learning Research 177:509-525 Available from https://proceedings.mlr.press/v177/lowe22a.html.

Related Material