Causal inference at the Max-Planck-Institute for Biological Cybernetics Tuebingen.

The research of our group aims at developing novel methods for inferring causal relations from data. For a detailed description see

Members of the causality group:

• Bernhard Schölkopf (MPI director)
• Dominik Janzing (senior research scientist, project leader)
• Joris Mooij (former research scientist)
• Kun Zhang (research scientist)
• David Balduzzi (research scientist)
• Kenji Fukumizu (research scientist)
• Jonas Peters (graduate student)
• Jakob Zscheischler (graduate student)
• Eleni Sgouritsa (graduate student)

Some publications and the available source code for them:

• Steudel, Janzing and Schölkopf (2010): Causal Markov condition for submodular information measures, code
• Zhang, Schölkopf and Janzing (2010): Invariant Gaussian Process Latent Variable Models and Application in Causal Discovery
• Peters, Janzing and Schölkopf (2010): Identifying Cause and Effect on Discrete Data using Additive Noise Models, code
• Peters, Janzing, Gretton and Schölkopf (2009): Detecting the Direction of Causal Time Series , code
• Janzing, Hoyer and Schölkopf (2010): Telling cause from effect based on high-dimensional observations, code
• Zhang and Hyvärinen (2009): Distinguishing cause from effect with constrained nonlinear ICA, code
• Hoyer, Janzing, Mooij, Peters, Schölkopf (2008): Nonlinear causal discovery with additive noise models, code
• Daniušis, Janzing, Mooij, Zscheischler, Steudel, Zhang, Schölkopf (2010): Inferring deterministic causal relations, code
• Mooij, Stegle, Janzing, Zhang, Schölkopf (2010): Probabilistic latent variable models for distinguishing between cause and effect, code
• Peters, Mooij, Janzing, Schölkopf (2011): Identifiability of Causal Graphs using Functional Models, code
• Zscheischler, Janzing, Zhang (2011): Testing whether linear relations are causal: A free probability approach, code
• Zhang, Peters, Janzing and Schölkopf (2011): Kernel-based conditional independence test and application in causal discovery, code
• Janzing, Sgouritsa, Stegle, Peters and Schölkopf (2011): Detecting low-complexity unobserved causes, code
• Mooij, Janzing, Heskes and Schölkopf (2011): On Causal Discovery with Cyclic Additive Noise Models, code

Cause-effect pairs dataset:

We are building a benchmark data set for causal discovery algorithms which focuses on the two-variable case. If you have interesting data that you would like to contribute please contact Jakob Zscheischler.