CHALLENGES IN OPTIMIZATION FOR DATA SCIENCE

Objective: In many data science problems, one is faced with large data sets, scarce or uncertain prior information, and restricted memory capabilities and computing power. Optimization – "the art of doing things better with limited resources" – naturally plays a central role in the various facets of data science. The aim of this conference is to bring together researchers and scientists with different background and expertise to discuss challenging issues in the modeling and the numerical solution of optimization problems arising in data science. The conference is organized around 15 interactive invited plenary talks.

Program: Titles disclosed at the beginning of each talk

• July 1, 2015
  
  • 08:50 – 09:00 : Opening address
  • 09:00 – 09:45 : V. Anantharam, Data-derived pointwise consistency slides
  • 09:45 – 10:30 : H. Attouch, Fast inertial dynamics for convex optimization – Convergence of FISTA algorithms slides
  • 10:30 – 11:00 : Break
  • 11:00 – 11:45 : F. Bach, Towards tighter convergence rates slides
  • 11:45 – 12:30 : S. Bubeck, Revisiting Nesterov's accelerated gradient descent method slides
  • 12:30 – 14:00 : Break
  • 14:00 – 14:45 : V. Chandrasekaran, Relative entropy optimization and applications slides
  • 14:45 – 15:30 : Y. LeCun, Large-scale optimization for deep learning slides
  • 15:30 – 16:00 : Break
  • 16:00 – 16:45 : E. De Vito, Kernel methods for support estimation slides
  • 16:45 – 17:30 : I. Giulini, A spectral clustering algorithm based on Gram operators slides
• July 2, 2015
  
  • 09:00 – 09:45 : E. Hazan, Projection-free optimization and learning slides
  • 09:45 – 10:30 : G. Iyengar, First-order algorithms for convex optimization slides
  • 10:30 – 11:00 : Break
  • 11:00 – 11:45 : L. Orecchia, Combinatorial meets convex slides
  • 11:45 – 12:30 : M. L. Overton, Nonsmooth, nonconvex optimization algorithms and examples slides
  • 12:30 – 14:00 : Break
  • 14:00 – 14:45 : N. Pustelnik, Some challenges in optimization and image reconstruction slides
  • 14:45 – 15:30 : J. W. Silverstein, Estimating population eigenvalues from large-dimensional sample covariance matrices slides
  • 15:30 – 16:00 : Break
  • 16:00 – 16:45 : S. J. Wright, Computations with coordinate descent methods slides

Organizers:

• Patrick Louis Combettes
  Laboratoire Jacques-Louis Lions
  Université Pierre et Marie Curie ‒ Paris 6, Paris, France
  Email
• Vianney Perchet
  Laboratoire de Probabilités et Modèles Aléatoires
  Université Paris Diderot ‒ Paris 7, Paris, France
  Email