SIAM Activity Group on Linear Algebra

In the context of the CIMI Semester on “High performance linear and nonlinear methods for large scale applications”, a workshop on “Innovative clustering methods for large graphsand block methods” will be held in Toulouse, France, from 6th to 8th July 2015.

The format that we plan for the workshop is for a few short presentations, on the topics of the day, leaving a lot of time for exchanges and discussions amongst participants. The central theme of the workshop is to investigate clustering techniques in the context of very large and sparse matrices, and the use of HPC to design efficient algorithms.

Over the three days of this workshop, the foci we plan for each day are: the first day will be on applications, describing various disciplines that can benefit from clustering, and what type of structure identification can be of interest depending on the application; this can for instance address the use of specific block structures in large linear system solvers, clustering in graphs with applications, and pattern recognition using numerical structure; the second day will focus on tools and kernels useful for clustering algorithms, such as scaling, graph reordering techniques, and spectral analysis; the last day will be devoted to tools for model reduction, including for instance non-negative factorization techniques and rank reduction techniques.

There are no registration fees for the workshop, but there is a limit on the number of participants that can be accommodated. Expressions of interest to attend or proposals (in the form of a one page abstract) for a short 20mn talk, should be sent by January 31st 2015, to either Daniel Ruiz or Sandrine Mouysset at IRIT ([email protected] or [email protected]), from whom further information can be obtained. Notifications of acceptance will be given by March 1st 2015.

A session of the conference “Sparse days in St Girons III” (29/06/2015-02/06/2015) will be dedicated to the topics of this workshop, and people interested in this area might also wish to attend the conference. Full details of how to register for the conference are available from the web page http://www.cimi-semester-hpc.fr.