John Langford @ eHarmony

http://photos2.meetupstatic.com/photos/event/8/a/9/a/event_158135482.jpeg

12:00 arrival and lunch served
12:30 John's talk
13:30 discussion

Title: Vowpal Wabbit, the Next Generation

Abstract: VW ( http://hunch.net/~vw ) is an ultrascale learning tool, capable of running at hardware speeds. I will discuss two new tricks of great use:

(1) Parallel learning over a cluster.

Using this, it is easy to train on terafeature size datasets with a kilocluster.

(2) Learning Reductions for complex problems.

Many complexprediction problems can be broken down into simple prediction problems. The reduction system in VW allows this to be done efficiently and effectively.

Bio:

John Langford studied Physics (http://en.wikipedia.org/wiki/Physics) and Computer Science (http://en.wikipedia.org/wiki/Computer_Science) at theCalifornia Institute of Technology (http://en.wikipedia.org/wiki/California_Institute_of_Technology), earning a double bachelor's degree in 1997, and received his Ph.D. (http://en.wikipedia.org/wiki/Ph.D.) from Carnegie Mellon University (http://en.wikipedia.org/wiki/Carnegie_Mellon_University) in 2002. Since then, he has worked at Yahoo!, Toyota Technological Institute, and IBM (http://en.wikipedia.org/wiki/IBM)'s Watson Research Center. He is also the primary author of the popular Machine Learning weblog (http://en.wikipedia.org/wiki/Weblog), hunch.net (http://www.hunch.net/) and the principle developer of Vowpal Wabbit (http://en.wikipedia.org/wiki/Vowpal_Wabbit). Previous research projects include Isomap, Captcha, Learning Reductions, Cover Trees, and Contextual Bandit learning. For more information visit http://hunch.net/~jl .

Lunch will be served

Bring your parking ticket for validation.