Amnon Shashua (School of Engineering and Computer Science, Hebrew University of Jerusalem)
Tuesday, 24.6.2008, 14:00
Quite a number of problems involving inference from data, whether visual data or otherwise, fall into the category of optimization. I will describe a general scheme for sequential and parallel (message passing) update rules based on the framework of Fenchel duality. The presentation will be done in the context of two applications (i) distributed
(parallel) Support-Vector-Machines (SVM) for handling large scale problems (to be presented at CVPR'08), (ii) Graphical Model inference over "convex free energies" providing a globally convergent message-passing algorithm with a similar architecture to Belief Propagation (to be presented at UAI'08).
Joint work with Tamir Hazan.