CS: Another postdoc, Spread spectrum MRI, Sparse Underwater Channel Estimation, Collaborative Filtering in a Non-Uniform World

Yves Wiaux just let me know of two items. The first one is a job for a postdoc in his lab, the text of the announcement is

BASP NODE ANNOUNCEMENT for a POSTDOCTORAL POSITION at LTS2:

Announcement February 2010.

In the context of new funding obtained by Dr Y. Wiaux at the Swiss National Science Foundation (SNSF) for the BASP research node, a two-year postdoctoral position is available as soon as in April 2010, on the theme "Compressed sensing imaging techniques for radio interferometry". This position will be hosted by the Signal Processing Laboratory 2 (LTS2) of the Signal Processing Laboratory (SP Lab) in the Institute of Electrical Engineering (IEL) of EPFL.

The position is opened to any dynamic and highly qualified candidate, Ph.D. in Electrical Engineering, Physics or equivalent and with a strong background in signal processing. Competence in programming (MATLAB, C) is required. Knowledge of compressed sensing, signal processing on the sphere, or radio-interferometric imaging is a plus. The successful candidate will be in charge of the collaborations between the BASP node and the international radio astronomy community. In addition to his/her main activities, he/she will also be welcome to collaborate with other researchers of the BASP node and of the SP Lab on signal processing for magnetic resonance imaging.

Note that EPFL offers very attractive salaries.

Requests for further information, and in a second stage applications, should be sent to Dr Y. Wiaux, directly by email.

I'll add it shortly to the compressive sensing jobs page. The second item is that the following paper reported on earlier:

"Spread spectrum for compressed sensing techniques in magnetic resonance imaging" by Y. Wiaux, G. Puy, R. Gruetter, J.-Ph. Thiran, D. Van de Ville, and P. Vandergheynst Preprint EPFL infoscience IEEE International Symp. on Biomedical Imaging: From Nano to macro (ISBI) (2010). IEEE Signal Process. Soc., in press

has been expanded in a journal paper:

Spread spectrum for accelerated acquisition in magnetic resonance imaging by Gilles Puy, Yves Wiaux, R. Gruetter, J-P. Thiran , D. Van De Ville, and Pierre Vandergheynst. The abstract reads:

We advocate the use of quadratic phase profiles in magnetic resonance imaging (MRI) with the aim of enhancing the achievable reconstruction quality from an incomplete k-space coverage for sparse or compressible signals, in the final perspective of accelerating the acquisition process relative to a standard complete coverage. The technique proposed amounts to the modulation of the image probed by a linear chirp. We analyze the related spread spectrum phenomenon in the context of the recent theory of compressed sensing, and we prove its effectiveness in enhancing the quality of image reconstruction, through a detailed analysis at each scale of a wavelet decomposition. We also establish the stability of this spread spectrum technique relative to noise, as well as its flexibility in terms of the chirp rate value required for optimal reconstruction. Our results rely both on theoretical considerations relative to the mutual coherence between the wavelet sparsity basis and the k-space sensing basis, as well as on extensive numerical simulations on the Shepp-Logan phantom.

The following two preprints just showed up on arxiv:

Compressed Sensing for Sparse Underwater Channel Estimation: Some Practical Considerations by Sushil Subramanian. the abstract reads:

We examine the use of a structured thresholding algorithm for sparse underwater channel estimation using compressed sensing. This method shows some improvements over standard algorithms for sparse channel estimation such as matching pursuit, iterative detection and least squares.

and Collaborative Filtering in a Non-Uniform World: Learning with the Weighted Trace Norm by Ruslan Salakhutdinov, Nathan Srebro. The abstract reads:

We show that matrix completion with trace-norm regularization can be significantly hurt when entries of the matrix are sampled non-uniformly. We introduce a weighted version of the trace-norm regularizer that works well also with non-uniform sampling. Our experimental results demonstrate that the weighted trace-norm regularization indeed yields significant gains on the (highly non-uniformly sampled) Netflix dataset.

Image credit: NASA/JPL/Space Science Institute, via the BadAstronomy blog