[Ml-stat-talks] Fwd: Reminder: CSML Seminar: Cynthia Rudin on Tuesday, October 20, 2015 at 12:30pm | Green Hall, Room 0-S-6

Barbara Engelhardt bee at princeton.edu
Mon Oct 19 09:11:06 EDT 2015


Talk of interest tomorrow.



Tuesday, October 20, 2015

Green Hall, Room 0-S-6

Cynthia Rudin- Massachusetts Institute of Technology



*Title*: “Thoughts on Decision Making using Multi-Armed Bandits and
Decision Trees”



*Abstract*: This talk is comprised of two mini-talks about ongoing
projects. This talk is a mix of storytelling, real-world multi-armed bandit
applications, and a sort of revival for decision trees.



* Mini-talk I: Regulating Greed Over Time for the Multi-Armed Bandit.



I will describe our high scoring entry in the Exploration-Exploitation 3
data mining competition. The goal of the competition was to build a better
recommender system Yahoo!'s Front Page, which provides personalized new
article recommendations. The main strategy we used was to carefully control
the balance between exploiting good articles and exploring new ones in the
multi-armed bandit setting. This strategy was based on an observation that
there were broad trends over time in the click-through-rates of the
articles. As it turns out, the types of broad trends we observe in the
Yahoo! data are in fact pervasive throughout many types of real data (well
beyond news articles). We thus formalize a setting where regulating greed
over time can be provably beneficial. This is captured through regret
bounds and leads to principled algorithms.



* Mini-talk II: Scalable Bayesian Rule Lists



I will discuss an approach to decision tree (rule list) learning that I
have been pursuing over the past few years. This method does not have the
disadvantage of greedy splitting and pruning that haunts decision tree
algorithms. It yields very sparse logical models in a computationally
efficient way. It is a fierce competitor for decision tree methods on a
wide variety of problems, and much more principled.



The work on multi-armed bandits is joint work with Stefano Traca, Ed Su,
and Ta Chiraphadhanakul. The work on decision lists is joint with Ben
Letham, David Madigan, Tyler McCormick, Hongyu Yang, and Margo Seltzer
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/ml-stat-talks/attachments/20151019/12198ce9/attachment.html>


More information about the Ml-stat-talks mailing list