Réunion

Inscriptions

28 août 2024 : la date limite d'inscription est atteinte, les inscriptions sont closes.

Scientific context

Numerous problems in signal/image processing, statistics, and machine learning rely on the resolution of sparse optimization problems. These problems are very challenging to solve due to their combinatorial nature. The purpose of this thematic day is to bring together researchers in the field in order to explore the recent progress made in solving these challenging non-convex optimization problems. In particular, the program will cover a variety of methods from continuous relaxation and non-convex optimization algorithms to global optimization techniques.

Organizers

L. Calatroni (CNRS, Laboratoire I3S), E. Soubies (CNRS, IRIT)

Program

9h - 9h10: Welcome

9h10-10h10: Jerome Idier (CNRS) + Laure Blanc-Féraud (CNRS): An overview of l_0 minimization

10h10h40: Coffee break

Session 1: Relaxations

10h40-11h20: Marcus Carlsson (Lund University, Sweden) On quadratic envelopes for low rank and sparsity problems

11h20-11h40: Paul Zheng (RWTH Aachen University, Germany - previously CVN-OPIS Inria CentraleSupelec), PENDANTSS: PEnalized Norm-ratios Disentangling Additive Noise, Trend and Sparse Spikes

11h40-12h00: Michel De Lara (Cermics, École des Ponts ParisTech) Hidden convexity in the l0-pseudo norm

12h-13h40: Lunch break and poster session

- M'hamed Essafri (INP Toulouse) On l_0 Bregman relaxations for Kullback-Leibler sparse regression

- Joseph Gabet (L2S CentraleSupélec) Global optimization in the context of spike deconvolution

- Anne Gagneux (ENS Lyon) Automated and Unbiased Coefficient Clustering with Non Convex SLOPE

- Seta Rakotomandimby (Cermics - ENPC) Abstract Cutting Plane Method applied to Sparse Optimization

Session 2: Optimization

13h40-14h20: Alain Rakotomamonjy (Criteo AI labs & Université de Rouen) Screening rule and working set algorithms for LASSO with non-convex regularizers.

14h20-14h40: Mohamed Mimoun (Aix-Marseille Université) Straight-Through meets Sparse Recovery: the Support Exploration Algorithm

14h40-15h00: Clara Lage (ENS Lyon) Identifying a piecewise affine signal from its nonlinear observation - application to DNA replication analysis

15h-00-15h30: Coffee break

Session 3: Global optimization

15h30:16h10: Sebastien Bourguignon (Ecole Centrale de Nantes) Branch-and-bound methods for sparse optimization: main topics and recent achievements

16h10-16h30: Theo Guyard (Inria and Insa Rennes) Toward stronger relaxations for L0-regularized problems

16h30-16h50: Diego Delle Donne (ESSEC Business School) A Novel Integer Linear Programming Approach for Global L0 Minimization

16h50-17h10: Nicolas Nadisic (Ghent University and Royal Institute for Cultural Heritage, Belgique) Global optimization for L0-constrained simultaneous sparse coding

17h10: Conclusions

Registration deadline: August 27 2024.

Registration is free but mandatory for organization purposes and maximal room capacity.