Talk 4.2: 1:30PM–2:45PM
| Session Name: | Signal Processing and Network Applications |
| Session Time: | 1:30PM–2:45PM |
| Author Name: | Delbert Dueck |
| Author Email: | delbert@psi.toronto.edu |
| Talk Title: | Affinity Propagation: Clustering by Passing Messages Between Data Points |
| Slides: | 4-2.ppt |
| Abstract: | Clustering data by identifying a subset of representative examples is important for processing sensory signals and detecting patterns in data. Such “exemplars” can be found by randomly choosing an initial subset of data points and then iteratively refining it, but this works well only if that initial choice is close to a good solution. In this talk, I will describe a new clustering method called “affinity propagation” that approximately considers all possible exemplar subsets, exchanging real-valued messages between data points until a high-quality solution gradually emerges. Affinity propagation has been used to cluster images of faces, detect genes in microarray data, identify cities efficiently accessed by airline travel, and summarize documents by clustering sentences. It has uniformly found clusters with much lower error than those found by other methods, and does so in less than one-hundredth the amount of time. Because of its simplicity, general applicability, and performance, I believe affinity propagation will prove to be of broad value in science and engineering. |
| Research Group: | Communications |
| Degree Program: | Ph.D. |
| Author Bio: | Delbert Dueck graduated with a B.Sc. in Computer Engineering from the University of Manitoba in 2002. He is presently working on a Ph.D. in Electrical Engineering at the University of Toronto in Brendan Frey's Probabilistic and Statistical Inference research lab. He was an intern with the Machine Learning and Applied Statistics group at Microsoft Research in Redmond, Washington for 2005 and 2006. Research interests include: machine learning and graphical models with applications in bioinformatics, communications, and computer vision. |