GT Statistique et imagerie
  • Présentation
  • Prochaines séances
  • Séances passées
  • Comment venir
  • Jeudi 25/06/2015  (salle A305) : GT special 
  • 10:00 - 11:00 : Anna Gilbert (University of Michigan) 
    Title: Sparse Approximation, List Decoding, and Uncertainty Principles
    Abstract: We consider list versions of sparse approximation problems, where unlike the existing results in sparse approxi- mation that consider situations with unique solutions, we are interested in multiple solutions. We introduce these problems and present the first combinatorial results on the output list size. These generalize and enhance some of the existing results on threshold phenomenon and uncertainty principles in sparse approximations. Our definitions and results are inspired by similar results in list decoding. We also present lower bound examples that bolster our results and show they are of the appropriate size. Joint work with Atri Rudra, Hung Ngo, Mahmoud Abo Khamis
  • 11:00 - 12:00 : short talks by students
    Jingwei Liang: Local Linear Convergence of Proximal Splitting Methods
    Quentin Denoyelle: support recovery of sparse deconvolution of positive measures
    Bernhard Schmitzer: Graph Metric Learning for Optimal Transport
  • 12:00 - 14:00 : lunch 
  • 14:00 - 15:00 : Anuj Srivastava (Florida Univ)
    Title : Phase-Amplitude Separation in Functional Data
    Abstract: Functional data is ubiquitous in all sciences, including computer vision where it is comes in form of real-valued functions, curves, surfaces, and images. A statistical analysis and modeling of functional data is often more natural via its decomposition into amplitude and phase components. Amplitude (or shape) captures the vertical variability in a function while phase represents a deformation of the domain. One can use either a statistical model or a proper Riemannian metric to define and extract these components from given data. We will present a framework based on elastic. Riemannian metric, with appropriate invariance properties, for this separation. Once separated, these components can be analyzed and modeled  using standard tools from  functional data analysis. An important strength of these methods is that objects (functions,  shapes, or images) are both registered and compared jointly, under the same criterion. The key idea in this framework is a family of square-root representations that transforms complicated elastic metrics to standard Euclidean metrics, resulting in a significant computational  simplification. I will demonstrate these ideas using simple examples from different application domains.
  • 15:00 - 16:00 : short talks by students : 
    Dario Prandi : A variational take on the sub-Riemannian model for the primary visual cortex and its applications to image processing
    Lenaic Chizat : An Interpolating Distance between Optimal Transport and Fisher-Rao
    Jonathan Vacher: Dynamic Texture Synthesis for Probing Motion Perception
Propulsé par Créez votre propre site Web unique avec des modèles personnalisables.