Aims of the iTWIST workshop series

The advent of increased computing capabilities, along with recent theoretical and numerical breakthroughs in the fields of signal processing, computational harmonic analysis, inverse problem solving, high-dimensional statistics and convex optimization, have boosted interactions between low-complexity data models (e.g., sparse or low-rank data models) and novel data sensing techniques.

In a nutshell, low-complexity data models aim at capturing, modeling and exploiting “just the information you need” in the ubiquitous data deluge characterizing any scientific or technological achievements. High dimensional objects can be thus reconstructed using little information. However, further developments and novel ideas are still required to meet new challenges, especially for efficiently dealing with complex data structures of "real life" applications and for interconnecting such models with other theoretical and applied fields.

​The iTWIST workshop aims at fostering collaboration between international scientific teams for developing new theories, applications and generalizations of low-complexity models. This is why this event emphasizes dissemination of ideas through both specific oral/poster presentations and free discussions.

For this edition, ITWIST will be divided into two parts, a 2-day doctoral school, followed by the actual workshop for three days.

.: Call for papers | Call for doctoral school participations :.

Topics of the workshop

For his 4th edition, iTWIST’18 will feature 8 invited talks (see below) and will have slots for about 40 oral and poster presentation. The workshop therefore welcomes 2-page paper submissions on any aspects of the following themes :

• Application of low-complexity models in non-convex/non-linear inverse problems such as phase retrieval, blind deconvolution, blind calibration, or dictionary learning.
• Machine learning, deep learning, compressive statistical learning and generalized high-dimensional inference of low-complexity statistical models.
• Optimization methods for recovering low-complexity data representations (e.g., low-rank or tensor decomposition).
• Information theory, high-dimensional (convex) geometry and randomness.
• Novel definition of low-complexity models (discrete-valued signals, co-sparsity, mixed/group norm, model-based, tensor/manifold models).
• Sensing and processing of low-complexity signals on manifolds and graphs.
• Complexity and accuracy tradeoffs in numerical methods/optimization.
• Simulation and/or developments of novel electronic/optical compressive sensors.

Plenary speakers:

(See this page for the description of the plenary lectures)

• Laurent Daudet (Institut Langevin, Paris)
• Mike Davies (University of Edinburgh, UK)
• Simon Foucart (Texas A&M University, USA)
• Deanna Needell (UCLA, USA)
• Gitta Kutyniok (TU Berlin, Germany)
• Ben Adcock (Simon Fraser University, Canada)
• Sjoerd Dirksen (RWTH Aachen, Germany)
• Ulugbek Kamilov (Washington University, USA)

Previous editions:

• iTWIST'16, Aalborg, Denmark
• iTWIST'14, Namur, Belgium
• iTWIST'12, Marseille, France

Next edition:

• iTWIST'20, Nantes, France