Groupe de Travail - Sequential Structured Statistical Learning

Wednesday May 17th, 2017
IHES, amphithéâtre Léon Motchane
35, route de Chartres - 91440 Bures-sur-Yvette



A number of modern applications of sequential decision making require developing strategies that are adaptive to the underlying structure of the data (graph, network, etc) and robust to change of the signal. This includes recommender systems incorporating social network information, cognitive radios, decentralized decision making, or robust structured reinforcement learning, to cite a few. In order to anticipate and impact the next generation of applications of the field, we thus need to push theory of sequential decision making to the next level, including recent development of spectral methods, low-rank matrices, and graph-based decision making. We intend to build visibility on this cross-disciplinary topic thanks to the organization of this seminar open to researchers from the field, from related fields as well as young researchers and PhD students.

Motivation, objectives

We intend to cover different fields of statistical machine learning, structured learning and sequential decision making addressing the following (non-exhaustive) list of questions and topics:
• Sequential data, time series, non-stationarity
• Statistical learning on graphs, networks.
• Low rank matrices, random graphs and random matrices.
• Spectral methods, latent variables.
• Dependent variables.
• Concentration of measure.
• Sequential decision making (e.g.,bandits, online learning, active learning), Reinforcement learning,
• Hidden Markov Models, Predictive State Representations, Feature Reinforcement Learning, Partially Observable Markov Decision Processes.
• Other topics whose relevance to the work group is well supported.

Relevance to the community

Solutions to such challenges will benefit the machine learning community at large, since they also appear in many real-world applications.


Registration is free but mandatory. Please fill the following form:

Invited Speakers

  • Olivier Catoni
  • Thomas Bonald
  • Pierre Alquier (or The Tien Mai)
  • Adrien Saumard
  • Paul Lagree
  • Rémy Degenne
  • Gilles Stoltz
  • Yohann Decastro


  • Odalric-Ambrym Maillard, CR INRIA, team SequeL, Lille.
  • Richard Combes, richard.combes@supé, Associate Professor, Centrale Supélec
  • Kinda Khawam,, Associate Professor, UVSQ, LRI


 Société Française de Statistique