High-dimensional Inference by Convex Optimisation


Our focus is on new theoretically and computationally solid techniques for a wide class of statistical problems through a new “operational” framework for analysis of inferential routines which relies on extensive use of the methodology involved in developing modern optimization.

Our research agenda includes, but does not reduce to
  • design of efficient testing routines with applications to diagnostics from heterogeneous data

  • new methods of recovery of signals from nonlinear observations with application in classification and identification and inverse problems for non-linear PDE

  • stochastic optimization utilizing indirect observations (e.g., privacy protected data)

  • design of robust procedures with focus on iterative (adversarial) adaptive techniques

  • large-scale online algorithmic implementation of inference routines

  • adaptive techniques for network inference from multi-sensor data, application to identification of temporal dynamics in biomedical data, (stochastic) optimization methods which allow efficient use of streaming data, …


The chair is collaborating with French companies Biomerieux and ST Microelectronics through supervision of 2 CIFRE PhD’s. We continue established collaborations with “MAGNET“ chair (by joint membership of E. Devijver), and chairs “Towards More Data Efficiency in Machine Learning” and “Optimization & Learning.” New collaborations are established with ENSAE-CREST through joint supervision of PhD projects.


A. Juditsky and A. Nemirovski published the monograph
Statistical Inference via Convex Optimization. Princeton University Press (2020) https://press.princeton.edu/books/hardcover/9780691197296/statistical-inference-via-convex-optimization

Talk at MIT ORC Seminar (online), Cambridge (2021)

Invited talk at MiMo Workshop (online) (2021)


Published on  January 11, 2024
Updated on January 11, 2024