Registration
Registration is free but mandatory for members of the SOCN partner institutions (UAntwerpen, VUB, ULB, UGent, KU Leuven, ULiège, UCLouvain, UMONS, UNamur). Please fill out the registration form at https://sites.uclouvain.be/socn/drupal/socn/study-day-registration.
Programme
8:30 Registration of participants and welcome coffee - hall Sainte-Barbe
9:50 Opening remarks - room BARB 93
10:00 Plenary lecture: ItsOPT: Inexact two-level smoothing optimization in the age of high-order regularization by Masoud Ahookhosh (UAntwerp) - room BARB 93
11:00 Poster spotlights, presentations of paper reviews, and presentations of open problems - room BARB 93
12:00 - 14:00 Coffee, poster session, lunch - hall Sainte-Barbe
14:00 Parallel sessions: paper reviews, open problems (or challenges), and short talks
1. Optimization and data science - Room BARB 93
- 14:00 [talk] Wim Vanroose (UAntwerpen), ResQPASS: an algorithm for bounded variable linear least squares with asymptotic Krylov convergence
- 14:30 [talk] Alireza Kabgani (UAntwerpen), High-order Moreau envelope in the nonconvex setting: Fundamental and differential properties
- 15:00 [talk] Florian Dubois (UMONS), An improved Sparse Subspace Clustering method using Homotopy
2. Matrix computations - Room BARB 94
- 14:00 [talk] Susan Ghaderi (KU Leuven / ESAT), Low-rank matrix factorization via Bregman gradient methods
- 14:30 [talk] Giovanni Barbarino (UMONS), Dual Simplex Volume Maximization for Simplex-Structured Matrix Factorization
- 15:00 [paper review] Sofiane Tanji (UCLouvain), Randomized Matrix Computations: Themes and Variations
3. Orthogonality - Room BARB 20
- 14:00 [talk] Simon Mataigne (UCLouvain), An efficient algorithm for the Riemannian logarithm on the Stiefel manifold for a family of Riemannian metrics
- 14:30 [open problem] Simon Mataigne (UCLouvain), An approximate solution to the matrix logarithm
- 15:00 [open problem] Florentin Goyens (UCLouvain), Approximate orthogonalization
15:30 Coffee - hall Sainte-Barbe
16:00 Plenary lecture: Passivity-based control in robotics: Virtual mechanisms, optimisation, and data-driven design by Fulvio Forni (University of Cambridge) - room BARB 93
17:00 End
Venue
The meeting will be held in the Auditoires UCLouvain Sainte-Barbe, Pl. Sainte-Barbe 1, 1348 Louvain-la-Neuve.
The venue is at walking distance (less than 10 minutes) from the Louvain-la-Neuve train station. If you come by car, the nearest free parking lots are "Parking Rédimé" and "Parking Baudouin 1er". If they are full, an alternative is it to use the paid parking lots in the city center, e.g., "Parking Grand-Rue".
Abstracts of the plenary lectures
ItsOPT: Inexact two-level smoothing optimization in the age of high-order regularization - Masoud Ahookhosh
In this talk, we introduce an inexact two-level optimization framework (ItsOPT) for finding first-order critical points of nonsmooth and nonconvex functions. The framework includes two levels: In the upper level, a smoothing technique (e.g., High-order Moreau envelope, high-order forward-backward envelope, high-order tensor envelope) will be applied to generate a smooth approximation of the objective function with the same minimizers. Then, first- or second-order methods will be introduced for minimizing the smoothing function. In the lower level, the corresponding high-order proximal subproblems (e.g., High-order proximal, forward-backward, and tensor subproblems) will be solved inexactly using subgradient or Bregman proximal methods. This will provide an approximated solution for the subproblems, leading to inexact smoothing information for the upper-level methods. We note that the complexity of solving the considered optimization problems is the multiplication of the complexities of both levels. Applying accelerated first- or second-order methods at the upper level and solving the subproblems with negligible complexity (e.g., logarithmic rate) may lead to the superfast methods attaining complexity better than worst-case complexity bounds. We finally introduce some algorithms and report preliminary numerical results.
Passivity-based control in robotics: virtual mechanisms, optimisation, and data-driven design - Fulvio Forni
Passivity-based control is a cornerstone of control theory and robotics (energy shaping and damping injection, control by interconnection). However, design and tuning of passive controllers remain challenging. In this talk we will discuss the core principles of passivity-based control and its application to contemporary robotics challenges. We will explore the idea of the feedback controller as a virtual mechanism, analogous to the physical morphology of a robot, whose structure and tuning shape the robot behaviour. To optimize controller performance, we will introduce algorithms for efficient gain tuning. Furthermore, we will combine constrained optimization with data-driven techniques for new passivity-based control methods tailored to real-world robotic applications.