Talk "Causality-inspired machine learning" by Uri Shalit, Technion, Haifa

We are happy to announce the following talk in the ETH Foundations of Data Science Seminar

Time: Wednesday, 22. January 2020, 13h15

Place: ETH Zurich HG E 1.1


Uri Shalit, Technion, Haifa, ”Causality-inspired machine learning"

We will present three recent projects where ideas from causal inference have inspired us to find new approaches to problems in machine learning. First, we will see how the idea of negative controls led us to find a way to perform off-policy evaluation in a partially observable Markov decision process (POMDP).

We will then present how using the idea of independence of cause and mechanism (ICM) can be used to help learn predictive models that are stable against a-priori unknown distributional shifts. Finally, we will see how thinking in terms of causal graphs led us to a new method for learning computer vision models that can better generalize to unseen object-attribute compositions in images.

Organisers: A. Bandeira, H. Bölcskei, P. Bühlmann, J. Buhmann, T. Hofmann, A. Krause, A. Lapidoth, H.-A. Loeliger, M. Maathuis, N. Meinshausen, G. Rätsch, C. Uhler, S. van de Geer, F. Yang

Seminar website:

Wed, Jan 22 at 11:00 - Talk by Guillaume Obozinski

The Data mining and machine learning group,, invites for the talk by Guillaume Obozinski on Wednesday, 22 January 2020, 11:00am-12:00am at Rue de la Tambourine 17, 1227 Carouge, Battelle building B, room 4.11, fourth floor


11:00 – 12:00

Guillaume Obozinski Deputy Chief Data Scientist at the Swiss Data Science Center

  • Convex unmixing and learning the effect of latent variables in Gaussian Graphical models with unobserved variables

    The edge structure of the graph defining an undirected graphical model describes precisely the structure of dependence between the variables in the graph. In many applications, the dependence structure is unknown and it is desirable to learn it from data, often because it is a preliminary step to be able to ascertain causal effects. This problem, known as structure learning, is hard in general, but for Gaussian graphical models it is slightly easier because the structure of the graph is given by the sparsity pattern of the precision matrix of the joint distribution, and because independence coincides with decorrelation. A major difficulty too often ignored in structure learning is the fact that if some variables are not observed, the marginal dependence graph over the observed variables will possibly be significantly more complex and no longer reflect the direct dependencies that are potentially associated with causal effects. In this work, we consider a family of latent variable Gaussian graphical models in which the graph of the joint distribution between observed and unobserved variables is sparse, and the unobserved variables are conditionally independent given the others. Prior work was able to recover the connectivity between observed variables, but could only identify the subspace spanned by unobserved variables, whereas we propose a convex optimization formulation based on structured matrix sparsity to estimate the complete connectivity of the complete graph including unobserved variables, given the knowledge of the number of missing variables, and a priori knowledge of their level of connectivity. Our formulation is supported by a theoretical result of identifiability of the latent dependence structure for sparse graphs in the infinite data limit, which is a particular instance of a more general result we prove for unmixing with convex norms. We propose an algorithm leveraging recent active set methods, which performs well in the experiments on synthetic data.

    Short Bio

    Guillaume Obozinski graduated with a PhD in Statistics from UC Berkeley in 2009. He did his postdoc and held until 2012 a researcher position in the Willow and Sierra teams at INRIA and Ecole Normale Supérieure in Paris. He was then Research Faculty at Ecole des Ponts ParisTech until 2018. Guillaume has broad interests in statistics and machine learning and worked over time on sparse modeling, optimization for large scale learning, graphical models, relational learning and semantic embeddings, with applications in various domains from computational biology to computer vision.

12:00 End

EU’s General Data Protection Regulation (GDPR) comes into force on 25th May 2018

The Swiss Statistical Society is committed to keeping only the minimum amount of data needed to stay in contact with its members and partners (name, address, email address and phone number), to not pass the data on to third parties and to immediately deleting all entries when requested.

Looking for a job?

Vacancies for statisticians and related professions are published here.

Calls for Papers

Calls for Papers/Contributions from scientific conferences and journals are published here as well as scientific competitions and awards.

List of Service Providers

Consulting firms and service providers for Statistics and Data Analytics are listed here.