Julien Mairal just sent me the following annoucement:

from the announcement:

Hi Igor,

here is a call for a post-doc for an ANR project.

http://lear.inrialpes.fr/people/mairal/resources/pdf/postdoc_macaron.pdf

When you have time, could you advertise it on your blog ? This is about local low-rank approximations for applications in bioinformatics and image processing. Thus, this would be a good match for nuit blanche !

Best regards.

from the announcement:

Research Topic and Objectives:

The goal of the MACARON project is to use data for solving scientific problems and automatically converting data into scientific knowledge by using machine learning techniques. We propose a research direction motivated by applications in bioinformatics and image processing. Low-rank matrix approximation is a popular tool for building web recommender systems [1] and plays an important role in large-scale classification problems in computer vision [2]. In many applications, we need however a different point of view. Data matrices are not exactly low-rank, but admit local low-rank structures [3]. This shift of paradigm is expected to achieve groundbreaking improvements over the classical low-rank paradigm, but it raises significant challenges that should be solved during the post-doc. The first objective is to develop new methodological tools to efficiently learn local low-rank structures in data. This will require both modeling skills (designing the right model) and good knowledge of optimization techniques (for efficient learning). The second objective is to adaptthese tools to genomic imputation problems and inverse problems in image processing

**Join the CompressiveSensing subreddit or the Google+ Community and post there !**

Liked this entry ? subscribe to Nuit Blanche's feed, there's more where that came from. You can also subscribe to Nuit Blanche by Email, explore the Big Picture in Compressive Sensing or the Matrix Factorization Jungle and join the conversations on compressive sensing, advanced matrix factorization and calibration issues on Linkedin.

## No comments:

Post a Comment