Publications

Distributional robustness of K-class estimators and the PULSE

Published in The Econometrics Journal, 2022

We show that the well-known K-class estimators possess interesting distributional robustness properties for out-of-distribution prediction. We propose a novel linear causal effect estimator (PULSE) motivated as the best predictive method among all methods that can not be rejected as being causal.

Recommended citation: Martin Emil Jakobsen and Jonas Peters (2022). "Distributional robustness of K-class estimators and the PULSE" The Econometrics Journal, 25(2), 404-432 https://doi.org/10.1093/ectj/utab031

Structure Learning for Directed Trees

Published in Journal of Machine Learning Research (JMLR), 2022

We propose a method to consistently estimate the underlying causal structure of non-linear additive noise directed tree models. Furthermore, we propose a procedure to test causal substructure hypotheses.

Recommended citation: Martin Emil Jakobsen, Rajen D. Shah, Peter Bühlmann and Jonas Peters (2022). "Structure Learning for Directed Trees", Journal of Machine Learning Research (JMLR). https://arxiv.org/abs/2108.08871

A causal framework for distribution generalization

Published in IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2021

We investigate the problem of out-of-distribution prediction from a causal model perspective where perturbations of the test-data distribution arise from interventions. We analyze the connection between the best predictive model and causal model. Finally, we propose a non-parametric causal effect estimator

Recommended citation: Rune Christiansen, Niklas Pfister, Martin Emil Jakobsen, Nicola Gnecco, and Jonas Peters (2021). "A causal framework for distribution generalization" IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), forthcoming. https://doi.org/10.1109/TPAMI.2021.3094760

Causal structure learning from time series: Large regression coefficients may predict causal links better in practice than small p-values

Published in Proceedings of the NeurIPS 2019 Competition and Demonstration Track, Proceedings of Machine Learning Research (PMLR), 2020

We describe and justify our time series structure learning algorithms that won the Causality 4 Climate competition at NeurIPS 2019.

Recommended citation: Sebastian Weichwald, Martin E. Jakobsen, Phillip B. Mogensen, Lasse Petersen, Nikolaj Thams, and Gherardo Varando (2020). Proceedings of the NeurIPS 2019 Competition and Demonstration Track, PMLR 123:27-36. http://proceedings.mlr.press/v123/weichwald20a/weichwald20a.pdf