Machine Learning Meetings and Events
Group Meetings: Group meetings are held Mondays from 11am to Noon (talk starts 11:10am) in D.L. Pratt 290C unless otherwise noted. Meetings are coordinated by Hugo Larochelle.
Tea Talks: Tea talks are held every Wednesday at 4:00pm in D.L. Pratt 290C. Talks should be simple, accessible, and not exceed 15 minutes. Speakers bring snacks, make tea, and provide a copy of the presented paper.
Tea Talk Oct 21, 2009: Max Margin Restricted Boltzmann Machines.
- Speaker: Ben Marlin
- Abstract: In this talk I will describe preliminary work on developing a maximum margin training algorithm for binary restricted Boltzmann machines. The training algorithm alternates between optimizing the model parameters given a set of fantasy data vectors, and selecting new fantasy data vectors with the highest probability under the current model parameters. The training algorithm thus replaces the need to sample from the current model distribution with the need to find the data configuration with the highest probability. We show that the problem of finding the data configuration with the highest probability involves a concave objective function defined over a non-convex set. We suggest a continuous relaxation and local search procedure for for obtaining an approximate solution to this sub-problem. We discuss properties of the model and training algorithm including a support vector-like property involving training/fantasy case pairs.