Topological methods for large and complex data sets

Monday, March 26, 2012 - 4:45pm - 5:30pm
Keller 3-180
Gunnar Carlsson (Stanford University)
Over the last ten years, a number of methodologies have been developed which leverage topological techniques and ways of thinking to provide understanding of point cloud data. These include ways of measuring shape via homological signatures, topological mapping techniques, and applications of certain kinds of diagram constructions to collections of samples. I will survey these methods, and propose some direct connections of these ideas with mainstream machine learning.
MSC Code: