Conference on information representation and estimation
University College London, London, UK.
September 6-8, 2010
Abstract:
Mathematical methods for signal processing and in general for data representation and inference are growing more and more sophisticated. Successful applications of such methods range from medical imaging to security. Developments in mathematics and signal processing are however often divergent. Therefore, the main aim of this conference is to bring together signal processing researchers with statisticians and mathematicians working on problems related to data modelling and estimation, to encourage the exchange of ideas as well as to discuss theoretical underpinning of the methods and domains of application.
The INSPIRE 2010 conference will be held at the Anatomy JZ Young LT at University College London from September 6 till September 8, 2010. So please, book these dates!
The conference includes two plenary talks and a few focused sessions. The plenary speakers are Prof. V. Goyal from Massachusetts Institute of Technology (MIT) and Prof. K Oweiss from Michigan State University. The focused sessions this year are on topics related to sparse inference, overcomplete representations and frames, climate and inference, signal processing in neuroscience, and machine learning.There will also be contributed session and posters. The contributed papers and posters are solicited in any area related to data representation and inference, but contributions covering the specific topics of the focused sessions are particularly welcome. We invite you to submit two-page extended abstracts, with pointers to reference material where appropriate. Submissions should be sent to p.dragottiATimperial.acDOTuk and should be received by June 15th 2010. Notification of acceptance will be given by July 10th 2010. Finally, there is going to be a tutorial session on methods of analysis in compressed sensing. Instructor: Dr Jared Tanner, EdinburghUniversity
Please notice that registration is free and includes lunches and coffee-breaks for the duration of the conference. Those contributing a paper or a poster will have the accommodation provided for the whole duration of the conference.
Conference on information representation and estimation
Electrical and Electronic Engineering Department,
Imperial College London, London, UK.
September 21-24, 2009
Abstract:
Mathematical methods for signal processing and in general for data representation and inference are growing more and more sophisticated. Successful applications of such methods range from medical imaging to security. Developments in mathematics and signal processing are however often divergent. Therefore, the main aim of this conference is to bring together signal processing researchers with statisticians and mathematicians working on problems related to data modelling and estimation, to encourage the exchange of ideas as well as to discuss theoretical underpinning of the methods and domains of application. The workshop will include three plenary talks, special sessions on focused topics, a poster session and tutorials. Professor Eric Moulines, Professor Zoubin Ghahramani and Professor Martin Vetterli are the confirmed plenary speakers.
There will be a banquet on Wednesday the 23th. Registration is free and those contributing to the conference will have the accommodation provided for the whole duration of the conference.
The aim of this workshop will be to draw together much of the recent work on algorithms which encourage sparsity, such as the minimisation of cost functions involving Lp-norms (typically 0 <= p < 2), with good methods for solving inverse problems on large datasets such as high-resolution images and 3D data. Usually such problems must be solved iteratively and there is a great need to ensure rapid convergence if the dataset is large, in order to avoid long computation times. Of particular interest are a number of recent papers on fast solutions to L1-minimisation problems and also on iterative threshold reduction methods that allow good solutions to be found to the non-convex L0-minimisation problem. Within the iterative context, it is also possible to adjust the weighting functions for terms in an L2-minimisation so that it approximates an L1 or L0 minimisation process.
In addition to well-known applications such as image deconvolution, there are strong links between this work and the emerging field of compressed sensing. The proposed workshop will discuss the above problem areas and attempt to unify the fairly diverse set of techniques that are currently being used into a more fundamental framework.
Technical Program
Day 1 (Sunday 14/12/08):
Coffee and registration from 10:00
11.15-11:30 Introduction by the organizers
11.30-12.30 Plenary talk 1: Remi Gribonval - Some stories about Lp-minimization, the
Restricted Isometry Property, and excessive pessimism.
Lunch
2:00-3:30 Oral Session
2:00-2:30 N. Kingsbury - Iterative Sparsity Methods for Coding and Deconvolution with
Overcomplete Transforms
2:30-3:00 P. Vandergheynst - Joint sparsity for multimodal signals
3:00-3:30 M. Plumbley - Stagewise Polytope Faces Pursuit for Recovery of Sparse
Representations
3.30-4.30 Tea and poster session (Posters 1-3-5-7-9-11)
4:30-6:00 Oral Session
4:30-5:00 L. Daudet - Divide and conquer: a few tactical approaches for on-the-fly /
parallelized sparse decompositions.
5:00-5:30 T. Blumensath - Generalising Sparsity: A Union of Subspaces Model
5:30-6:00 Y. Eldar - Compressed Sensing of Analog Signals
