The embattled affinity propagation algorithm and its applications in
                biomedicine and bioinformatics

                        Brendan J. Frey
                     University of Toronto

Data sets arising in biomedicine and bioinformatics are often huge and
consist of quite different types of data (eg, sequence data and microarray
measurements). Consequently, standard machine learning techniques usually
cannot be directly applied. In this talk, I'll describe an algorithm
called affinity propagation and discuss why it offers flexibility in
analyzing the kinds of data sets arising in bioinformatics and
biomedicine. I'll describe applications in the areas of whole-genome
transcript detection using microarrays, image segmentation, text analysis
and motif discovery. Affinity propagation can implemented in a couple
dozen lines of MATLAB or C and is suitable for distributed computing
environments, making it attractive for high-throughput computations.