Divide and Conquer Techniques for Machine Learning

Picture of Mike Izbicki
Speaker Name: 
Mike Izbicki
Speaker Title: 
PhD Student
Speaker Organization: 
UC Riverside
Start Time: 
Tuesday, June 13, 2017 - 2:00pm
End Time: 
Tuesday, June 13, 2017 - 3:00pm
UCSC, Engineering 2, room 599
Lise Getoor

Abstract: Two algorithms will be presented that use divide and conquer techniques to speed up learning.  The first algorithm (called OWA = optimal weighted average) is a communication efficient distributed learner.  OWA requires only a single round of communication, which is sufficient to achieve optimal learning rates.  The second algorithm is a meta-algorithm for fast cross validation.  It will show that for any divide and conquer learning algorithm, there exists a fast cross validation procedure whose run time is asymptotically independent of the number of cross validation folds.

Bio: Mike Izbicki is a Ph.D. student at UC Riverside.  His research uses techniques from high dimensional statistics to make machine learning algorithms faster and easier to use.  Mike has also done a lot of teaching abroad and advocating for open source software.