MUSCLE - E-team 13 Final Activity Report

Salerno E.

Machine learning  Blind source separation  Soft segmentation  MUSCLE-NoE 

This report describes the common activities performed within the MUSCLE e-team on unsupervised segmentation and classification of multichannel data. Active under MUSCLE JPA3 and JPA4, this e-team has been first named e-team 15 and then (after June 2007) e-team 13. The work plan included the implementation of tools for remote collaboration, integration of research on statistical approaches to soft segmentation, and organization of both internal and open meetings to foster collaboration and disseminate our work. The strict collaboration on Bayesian separation techniques has produced a paper, ready to be submitted for publication, on a promising MCMC technique to separate astrophysical source maps, which is being assessed with the help of astrophysicists. Further work on document image processing and remote sensing has also produced results documented in the literature. The e-team organized an invited session on computational learning for unsupervised segmentation at the international conference KES 2007.

Source: Project report, MUSCLE, Deliverable DN6.1, 2008