Julius von Kügelgen

Exploring the intersection of causal inference and machine learning


I am a postdoc with Jonas Peters at the Seminar for Statistics at ETH Zürich, and an associated researcher at the ETH AI Center. My research interests lie at the intersection of causal inference and machine learning, see Research Interests and my publications for more details. 

I obtained my PhD in the Cambridge-Tübingen programme, co-advised by Bernhard Schölkopf at the Max Planck Institute for Intelligent Systems and Adrian Weller at the University of Cambridge. During my PhD, I interned at Amazon and visited Columbia University (working with Dave Blei & Elias Bareinboim) and UC Berkeley (attending the Simons Institute Causality Program). 

Previously, I studied Mathematics (B.Sc., M.Sci.) at Imperial College London (UK) and Artificial Intelligence (M.Sc.) at UPC Barcelona (Spain) and at TU Delft (Netherlands).


News & Updates

PhD Thesis

Identifiable Causal Representation Learning:
Unsupervised, Multi-View, and Multi-Environment

Julius von Kügelgen

[Link to PDF]

Highlighted Publications

Multi-View Causal Representation Learning with Partial Observability.

Dingling Yao, Danru Xu, Sebastien Lachapelle, Sara Magliacane, Perouz Taslakian, Georg Martius, JvK, Francesco Locatello

ICLR 2024 (Spotlight; top 3%)
(also at: Workshop on Causal Representation Learning @ NeurIPS 2023, Oral)


Self-Supervised Disentanglement by Leveraging Structure in Data Augmentations.

Cian Eastwood, JvK, Linus Ericsson, Diane Bouchacourt, Pascal Vincent, Bernhard Schölkopf, Mark Ibrahim

Workshop on Causal Representation Learning @ NeurIPS 2023


Nonparametric Identifiability of Causal Representations from Unknown Interventions.

JvK, Michel Besserve, Wendong Liang, Luigi Gresele, Armin Kekić, Elias Bareinboim, David M. Blei, Bernhard Schölkopf

NeurIPS 2023


Causal Component Analysis.

Wendong Liang, Armin Kekić, JvK, Simon Buchholz, Michel Besserve, Luigi Gresele, Bernhard Schölkopf

NeurIPS 2023


Spuriosity Didn’t Kill the Classifier: Using Invariant Predictions to Harness Spurious Features.

Cian Eastwood, Shashank Singh, Andrei L. Nicolicioiu, Marin Vlastelica, JvK, Bernhard Schölkopf

NeurIPS 2023
(also at ICML 2023 Workshop Spurious Correlations, Invariance and Stability)


Provably Learning Object-Centric Representations.

Jack Brady, Roland S. Zimmermann, Yash Sharma, Bernhard Schölkopf, JvK*, Wieland Brendel* (*shared last author)

ICML 2023 (Oral)

[arXiv] [project page]

Backtracking Counterfactuals.

JvK, Abdirisak Mohamed, Sander Beckers.

CLeaR 2023 (Best Paper Award)

[arXiv] [long presentation] [slides]

Active Bayesian Causal Inference.

Christian Toth, Lars Lorch, Christian Knoll, Andreas Krause, Franz Pernkopf, Robert Peharz*, JvK*. (*shared last author)

NeurIPS 2022
(also at: NeurIPS 2022 Workshops Causality for Real-world Impact & Neuro Causal and Symbolic AI)

[arXiv] [poster] [code]

Causal Discovery in Heterogeneous Environments Under the Sparse Mechanism Shift Hypothesis.

Ronan Perry, JvK*, Bernhard Schölkopf*. (*shared last author)

NeurIPS 2022

[arXiv] [poster] [code]

Probable Domain Generalization via Quantile Risk Minimization.

Cian Eastwood*, Alexander Robey*, Shashank Singh, JvK, Hamed Hassani, George J. Pappas, Bernhard Schölkopf.

NeurIPS 2022

[arXiv] [poster] [code]

Causal Inference Through the Structural Causal Marginal Problem.

Luigi Gresele*, JvK*, Jonas M. Kübler*, Elke Kirschbaum, Bernhard Schölkopf, Dominik Janzing. (*equal contribution)

ICML 2022


On the Fairness of Causal Algorithmic Recourse.

JvK, Amir-Hossein Karimi, Umang Bhatt, Isabel Valera, Adrian Weller, Bernhard Schölkopf.

AAAI 2022 (Oral)
(also at: ICML 2021 Workshop Algorithmic Recourse; NeurIPS 2020 Workshop Algorithmic Fairness through the Lens of Causality and Interpretability (AFCI) )

[paper] [code] [poster] [recording]

Self-Supervised Learning with Data Augmentations Provably Isolates Content from Style.

JvK*, Yash Sharma*, Luigi Gresele*, Wieland Brendel, Bernhard Schölkopf, Michel Besserve, Francesco Locatello. (*equal contribution)

NeurIPS 2021

[arXiv] [code] [recording] [long presentation] [slides]

Independent Mechanism Analysis, A New Concept?

Luigi Gresele*, JvK*, Vincent Stimper, Bernhard Schölkopf, Michel Besserve. (*equal contribution)

NeurIPS 2021

[arXiv] [code] [recording]

Algorithmic Recourse Under Imperfect Causal Knowledge: A Probabilistic Approach

Amir-Hossein Karimi*, JvK*, Bernhard Schölkopf, Isabel Valera. (*equal contribution)

NeurIPS 2020 (Spotlight)
(also at: ICML 2020 Workshops

[paper] [short presentation] [long presentation (@UCL reading group)] [poster]

Other Publications & Preprints

A Sparsity Principle for Partially Observable Causal Representation Learning.

Danru Xu, Dingling Yao, Sebastien Lachapelle, Perouz Taslakian, JvK, Francesco Locatello, Sara Magliacane.

ICML 2024
(Previously at: Workshop on Causal Representation Learning @ NeurIPS 2023)


Independent Mechanism Analysis and the Manifold Hypothesis.

Shubhangi Ghosh, Luigi Gresele, JvK, Michel Besserve, Bernhard Schölkopf.

Workshop on Causal Representation Learning @ NeurIPS 2023


Deep Backtracking Counterfactuals for Causally Compliant Explanations.

Klaus-Rudolf Kladny, JvK, Bernhard Schölkopf, Michael Muehlebach.


Kernel-Based Independence Tests for Causal Structure Learning on Functional Data

Felix Laumann, JvK, Junhyung Park, Bernhard Schölkopf, Mauricio Barahona

Entropy, 2023 (Special Issue on Causality and Complex Systems)

[OpenAccess] [arxiv]

Causal Effect Estimation from Observational and Interventional Data Through Matrix Weighted Linear Estimators.

Klaus-Rudolf Kladny, JvK, Bernhard Schölkopf, Michael Muehlebach.

UAI 2023


Evaluating Vaccine Allocation Strategies Using Simulation-Assisted Causal Modelling.

Armin Kekić, Jonas Dehning, Luigi Gresele, JvK, Viola Priesemann, Bernhard Schölkopf.

(Previously at: NeurIPS 2022 Workshop A Causal View on Dynamical Systems)

[arxiv] [pdf]

Unsupervised Object Learning via Common Fate.

Matthias Tangemann, Steffen Schneider, JvK, Francesco Locatello, Peter Gehler, Thomas Brox, Matthias Kümmerer, Matthias Bethge, Bernhard Schölkopf.

CLeaR 2023


DCI-ES: An Extended Disentanglement Framework with Connections to Identifiability.

Cian Eastwood*, Andrei Liviu Nicolicioiu*, JvK*, Armin Kekic, Frederik Träuble, Andrea Dittadi, Bernhard Schölkopf. (*equal contribution)

ICLR 2023
(Previously at: Workshop on Causal Representation Learning @ UAI 2022)


Embrace the Gap: VAEs Perform Independent Mechanism Analysis.

Patrik Reizinger, Luigi Gresele, Jack Brady, JvK, Dominik Zietlow, Bernhard Schölkopf, Georg Martius, Wieland Brendel, Michel Besserve.

NeurIPS 2022
(Previously at: 5th Workshop on Tractable Probabilistic Modeling @ UAI 2022)

[arXiv] [poster]

From Statistical to Causal Learning.

Bernhard Schölkopf*, JvK*. (*equal contribution)

Proceedings of the International Congress of Mathematicians 2022


Complex interlinkages, key objectives and nexuses amongst the Sustainable Development Goals and climate change: a network analysis

Felix Laumann, JvK, Thiago Hector Kanashiro Uehara, Mauricio Barahona

The Lancet Planetary Health, 2022


Towards Causal Algorithmic Recourse.

Amir-Hossein Karimi*, JvK*, Bernhard Schölkopf, Isabel Valera.

xxAI - Beyond Explainable AI---Lecture Notes in Computer Science, vol. 13200, 2022.


You Mostly Walk Alone: Analyzing Feature Attribution in Trajectory Prediction.

Osama Makansi, JvK, Francesco Locatello, Peter Gehler, Dominik Janzing, Thomas Brox, Bernhard Schölkopf.

ICLR 2022


Visual Representation Learning Does Not Generalize Strongly Within the Same Domain.

Lukas Schott, JvK, Frederik Träuble, Peter Gehler, Chris Russell, Matthias Bethge, Bernhard Schölkopf, Francesco Locatello, Wieland Brendel.

ICLR 2022
(Previously at: ICLR 2021 Workshop Generalization beyond the training distribution in brains and machines )

[paper[poster]  [code/benchmark]

Causal Direction of Data Collection Matters: Implications of Causal and Anticausal Learning for NLP.

Zhijing Jin*, JvK*, Jingwei Ni, Tejas Vaidhya, Ayush Kaushal, Mrinmaya Sachan, Bernhard Schölkopf. (*equal contribution)

EMNLP 2021 (Oral)

[arXiv] [code] [poster] [slides] [talk]

Backward-Compatible Prediction Updates: A Probabilistic Approach.

Frederik Träuble, JvK, Matthäus Kleindessner, Francesco Locatello, Bernhard Schölkopf, Peter Gehler.

NeurIPS 2021

[arXiv] [recording]

Simpson's paradox in Covid-19 case fatality rates: a mediation analysis of age-related causal effects.

JvK*, Luigi Gresele*, Bernhard Schölkopf. (*equal contribution)

IEEE Transactions on Artificial Intelligence, 2021.

[paper] [data & code] [video1] [video2]

Algorithmic recourse in partially and fully confounded settings through bounding counterfactual effects. 

JvK, Nikita Agarwal, Jakob Zeitler, Afsaneh Mastouri, Bernhard Schölkopf.

ICML 2021 Workshop Algorithmic Recourse


Kernel Two-Sample and Independence Tests for Non-Stationary Random Processes.

Felix Laumann, JvK, Mauricio Barahona.

7th International Conference on Time Series and Forecasting (ITISE 2021)


Towards Causal Generative Scene Models via Competition of Experts.

 JvK*, Ivan Ustyuzhaninov*, Peter Gehler, Matthias Bethge, Bernhard Schölkopf. (*equal contribution)

ICLR 2020 Workshop Causal Learning for Decision Making (CLDM)

[paper]  [presentation (video)]

Semi-supervised learning, causality and the conditional cluster assumption.

JvK, Alexander Mey, Marco Loog, Bernhard Schölkopf.

UAI 2020
(also at: NeurIPS 2019 Workshop “Do the right thing”: machine learning and causal inference for improved decision making)

[paper]  [poster]

Optimal experimental design via Bayesian optimisation: active causal structure learning for Gaussian process networks.

JvK, Paul K Rubenstein, Bernhard Schölkopf, Adrian Weller.

NeurIPS 2019 Workshop “Do the right thing”: machine learning and causal inference for improved decision making

[paper]  [poster]

Semi-generative modelling: covariate-shift adaptation with cause and effect features.

JvK, Alexander Mey, Marco Loog.

(Previously: ICML 2018 Workshop CausalML)