GdR MIA Thematic day on Non-Convex Sparse Optimization

Friday October 9th 2020 at ENSEEIHT, Toulouse


Sparse models are widely used in machine learning, statistics, and signal/image processing applications. They usually lead to NP-hard non-convex optimization problems that involve the l0 pseudo-norm. The purpose of this thematic day is to bring together speakers from different teams in order to explore the recent progress made in solving these challenging non-convex optimization problems. We expect to cover a variety of methods that include (but not only) greedy algorithms, continuous relaxations, screening rules, as well as global optimization through branch-and-bound strategies or semidefinite programming.


Full program with abstracts [here]

Chair Emmanuel Soubies
08h30 - 09h00 Manlio Gaudioso A k-norm-based Mixed Integer Programming formulation for sparse optimization.
09h00 - 09h30 Sébastien Bourguignon Exact l0-norm optimization via branch-and-bound methods.
09h30 - 10h00 Arthur Marmin Global solution to non-convex optimization problems involving an approximate l0 penalization.
10h00 - 10h30 ---- Coffee break ----
Chair Cássio Fraga Dantas
10h30 - 11h00 Vasiliki Stergiopoulou Microscopic super-resolution in the lateral plane by sparse regularization in the correlation domain.
11h00 - 11h30 Fiorella Sgallari Convex non-convex variational models for image processing.
11h30 - 12h00 Joseph Salmon Safe screening and active sets for sparse regularization.
12h00 - 14h00 ---- Lunch ----
Chair Cédric Févotte
14h00 - 14h30 Charles Soussen Exact recovery analysis of non-negative orthogonal greedy algorithms.
14h30 - 15h00 Liva Ralaivola Recovery and convergence rate of the Frank-Wolfe algorithm for the m-exact-sparse problem.
15h00 - 15h30 Yann Traonmilin A framework for non-convex recovery of low dimensional models in infinite dimension.
15h30 - 16h00 ---- Coffee break ----
Chair Henrique Goulart
16h00 - 16h30 Léon Zheng Identifiability in matrix sparse factorization.
16h30 - 17h00 Nicolas Nadisic Sparse separable nonnegative matrix factorization.