Senior Research Scientist

Office: 208

Spemannstr. 38

72076 Tübingen

Spemannstr. 38

72076 Tübingen

+49 7071 601-563

+49 7071 601 552

There has been a long history of debate on causality in philosophy, statistics, economics, and related fields. I have been concerned with this classic question--how can we discover causal information from purely observed data (i.e., perform **causal inference**)? How such causal information can facilitate solving other problems such as modeling, prediction, and control, is also interesting to me.

My research consists of three main lines.

- First, I have focused on developing
*practical computational methods for causal inference*, to produce more reliable causal information. - Secondly, to better understand causality and derive more universal methods for causal inference, I also work on finding
*fundamental and testable principles that help discover causality from data*. - Thirdly, latent variable modeling is closely related to causality, and it has been interesting me for over eight years. Developing more
*general yet identifiable latent variable*models would benefit the causality field, as well as the machine learning and signal processing communities.

Since machine learning plays a key role in data analysis as well as causal inference, I am also very interested in this field.

- The workshop "Causal modeling & machine learning" will take place in Beijing, China, in June 2014.
- We are editing the ACM Transactions on Intelligent Systems and Technologies (ACM TIST) special issue on causal discovery and inference; see the call for papers here. Submission deadline: 14 March 2014.
- The workshop "Causality: Perspectives from different disciplines" took place in August, 2013.
- Slides and poster for a recent paper "Domain adaptation under target and conditional shift."

- fundamental
**characterization**of causal information in observational data, and refinement of concepts related to causality - precise notion of “model
**complexity**” for causal inference - unified/universal
**approach**for causal inference **domain-specific**causal inference (in finance, brain signal analysis, etc.)- causal
**understanding**of machine learning tasks - practical causal inference system for
**large-scale**problems - domain
**adaptation** **big data**analytics: a causal perspective- computational
**finance**

- Causal discovery: Theory and applications
- developing advanced and practical computational methods for causal inference
- finding fundamental and testable principles to characterize causality
- latent variable modeling
- Statistical machine learning and applications
- kernel methods, Gaussian processes, domain adaptation, mixture models, model selection, independent component analysis, sparse coding
- Computational finance
- Neuroscience (especially MEG and EEG data analysis)

- Organizational activities
- Organizer of ICML'14 workshop "Causal modeling and machine learning" (with Bernhard Schölkopf, Elias Bareinboim, and Jiji Zhang), June, 2014
- Guest editor of ACM Transactions on Intelligent Systems and Technologie special issue on Causality (with Jiuyong Li, Elias Bareinboim, Bernhard Schölkopf, and Judea Pearl)
- Organizer of workshop "Causality: Perspectives from different disciplines" (with Bernhard Schölkopf and Jiji Zhang), Vals, Switzerland, August 5-8, 2013
- Co-organizer of the First IEEE ICDM Workshop on Causal Discovery (CD 2013), Dallas, Texas, USA, December 8, 2013
- Co-organizer of workshop “Networks -- Processes and causality”, Menorca, Spain, September, 2012
- Publicity chair of AISTATS 2012 (15th International Conference on Artificial Intelligence and Statistics)
- Reviewer for journals
- Annals of Statistics; Journal of Machine Learning Research; Annals of Applied Statistics; Journal of the American Statistical Association; Neural Computation; Machine Learning; IEEE Transactions on Pattern Analysis and Machine Intelligence; IEEE Transactions on Neural Networks; IEEE Transactions on Signal Processing; Neural Networks; IEEE Transactions on Knowledge and Data Engineering; Quantitative Finance; Neurocomputing; IEEE Signal Processing Letters; Frontiers of Computer Science; International Journal of Imaging Systems and Technology; Circuits, Systems & Signal Processing; International Review of Economics and Finance
- Program committee member for international conferences
- 2014: AISTATS (senior program committee), UAI, NIPS, WSDM, KDD, iKDD CoDS...
- 2013: UAI, NIPS, AISTATS, SDM, KDD, IJCAI, IJCNN, ASE/IEEE Big Data;
- 2012: UAI, AISTATS, MLSP, WSDM, SDM;
- 2011: UAI, NIPS, KDD, IJCNN, ICONIP;
- 2010: UAI, NIPS, ICA/LVA, SDM, ACML, ICPR;
- 2009: NIPS, ACML, ICONIP

49 results
(BibTeX)

**Recent Methodological Advances in Causal Discovery and Inference** In: *15th Conference on Theoretical Aspects of Rationality and Knowledge*, TARK 2015
(In Proceedings)

**Causal Inference by Identification of Vector Autoregressive Processes with Hidden Components** In: *Proceedings of the 32nd International Conference on Machine Learning*, 37, 1917–1925, JMLR, ICML 2015
(In Proceedings)

**Discovering Temporal Causal Relations from Subsampled Data** In: *Proceedings of the 32nd International Conference on Machine Learning*, 37, 1898–1906, JMLR, ICML 2015
(In Proceedings)

**Distinguishing Cause from Effect Based on Exogeneity** In: *Fifteenth Conference on Theoretical Aspects of Rationality and Knowledge*, TARK 2015
(In Proceedings)

**Identification of Time-Dependent Causal Model: A Gaussian Process Treatment** In: *24th International Joint Conference on Artificial Intelligence, Machine Learning Track*, IJCAI15
(In Proceedings)

**Multi-Source Domain Adaptation: A Causal View** In: *Proceedings of the Twenty-Ninth AAAI Conference on Artificial Intelligence*, 3150-3157, AAAI Press, AAAI 2015
(In Proceedings)

**A Permutation-Based Kernel Conditional Independence Test** In: *Proceedings of the 30th Conference on Uncertainty in Artificial Intelligence (UAI2014)*, 132–141, AUAI Press Corvallis, Oregon, UAI2014
(In Proceedings)

**Likelihood and Consilience: On Forster’s Counterexamples to the Likelihood Theory of Evidence** *Philosophy of Science, Supplementary Volume 2015*
(Article)

**Causal discovery via reproducing kernel Hilbert space embeddings** *Neural Computation*, 26(7):1484-1517
(Article)

**Single-Source Domain Adaptation with Target and Conditional Shift** In: *Regularization, Optimization, Kernels, and Support Vector Machines*, 427-456, Chapman and Hall/CRC, Boca Raton, USA
(Book Chapter)

**On estimation of functional causal models: General results and application to post-nonlinear causal model** *ACM Transactions on Intelligent Systems and Technologies*
(Article)

**Nonlinear Causal Discovery for High Dimensional Data: A Kernelized Trace Method** IEEE Computer Society, IEEE International Conference on Data Mining (ICDM’13)
(In Proceedings)

**Domain adaptation under Target and Conditional Shift** In: *Proceedings of the 30th International Conference on Machine Learning, W&CP 28 (3)*, 819–827, JMLR, ICML 2013
(In Proceedings)

**On estimation of functional causal models: Post-nonlinear causal model as an example** In: *First IEEE ICDM workshop on causal discovery *
(In Proceedings)

**Semi-supervised learning in causal and anticausal settings** In: *Empirical Inference*, 129–141, Springer-Verlag
(Book Chapter)

**On Causal and Anticausal Learning** In: *Proceedings of the 29th International Conference on Machine Learning (ICML)*, 1255-1262, Omnipress, New York, NY, USA, ICML 2012
(In Proceedings)

**Information-geometric approach to inferring causal directions** *Artificial Intelligence*, 182-183, 1-31
(Article)

**Causal discovery with scale-mixture model for spatiotemporal variance dependencies ** In: *Advances in Neural Information Processing Systems 25*, 1736–1744, Curran Associates Inc., 26th Annual Conference on Neural Information Processing Systems (NIPS 2012)
(In Proceedings)

**A general linear non-Gaussian state-space model: Identifiability, identification, and applications** In: *JMLR Workshop and Conference Proceedings Volume 20*, 113-128, MIT Press, Cambridge, MA, USA, 3rd Asian Conference on Machine Learning (ACML 2011)
(In Proceedings)

**Testing whether linear equations are causal: A free probability theory approach** 839-847, AUAI Press, Corvallis, OR, USA, 27th Conference on Uncertainty in Artificial Intelligence (UAI 2011)
(In Proceedings)

There has been a long history of debate on causality in philosophy, statistics, economics, and related fields. I have been concerned with this classic question--how can we discover causal information from purely observed data (i.e., perform **causal inference**)? How such causal information can facilitate solving other problems such as modeling, prediction, and control, is also interesting to me.

My research consists of three main lines.

- First, I have focused on developing
*practical computational methods for causal inference*, to produce more reliable causal information. - Secondly, to better understand causality and derive more universal methods for causal inference, I also work on finding
*fundamental and testable principles that help discover causality from data*. - Thirdly, latent variable modeling is closely related to causality, and it has been interesting me for over eight years. Developing more
*general yet identifiable latent variable*models would benefit the causality field, as well as the machine learning and signal processing communities.

Since machine learning plays a key role in data analysis as well as causal inference, I am also very interested in this field.

- The workshop "Causal modeling & machine learning" will take place in Beijing, China, in June 2014.
- We are editing the ACM Transactions on Intelligent Systems and Technologies (ACM TIST) special issue on causal discovery and inference; see the call for papers here. Submission deadline: 14 March 2014.
- The workshop "Causality: Perspectives from different disciplines" took place in August, 2013.
- Slides and poster for a recent paper "Domain adaptation under target and conditional shift."

Office: 208

Spemannstr. 38

72076 Tübingen

Spemannstr. 38

72076 Tübingen

+49 7071 601-563

+49 7071 601 552