Seminar Details
2005-11-08 (14.00 pm) : Non-Negative Matrix Factorizations and Graph Clustering
At Auditoire Euler
Organized by Mathematical Engineering
Speaker :
Diep Ho (CESAME)
Abstract :
We present a method to approximate a non-negative matrix A by a low-rank
product UV whose factors U and V are also non-negative.
This kind of approximation is useful in applications such as face learning
and recognition, and in clustering of non-negative data.
In particular, we will focus on the application of this method to the problem
of Graph Clustering where vertices of a graph are to be grouped
into clusters with high connectivity within each cluster and low connectivity
across clusters.
We present some promising results and mention some open problems.