Skip navigation
Please use this identifier to cite or link to this item:
Title: A Survey of Strategies for the Multi-Armed Bandit Problem
Authors: Korać, Damjan
Advisors: Bubeck, Sébastien
Department: Operations Research and Financial Engineering
Class Year: 2013
Abstract: We explore, in increasing complexity, various strategies for addressing the Multi- Armed Bandit Problem in order to find the algorithm that maximizes the clickthrough rate in the Exploration and Exploitation Challenge 3 competition. The data used result from the actions of web site visitors who clicked on news articles, and we utilize different policies to decide how to optimally display the various articles. We implement several policies and find a score for two versions of each algorithm: one that is primed on test data and one that encounters the actual competition data without any prior knowledge. An Upper Con dence Bound strategy proves to outperform other algorithms for the tuned trials, and a more recent policy with bounded regrets serves as the best out of the box strategy.
Extent: 84 pages
Access Restrictions: Walk-in Access. This thesis can only be viewed on computer terminals at the Mudd Manuscript Library.
Type of Material: Princeton University Senior Theses
Language: en_US
Appears in Collections:Operations Research and Financial Engineering, 2000-2016

Files in This Item:
File SizeFormat 
Korac Damjan Thesis.pdf1.85 MBAdobe PDF    Request a copy

Items in Dataspace are protected by copyright, with all rights reserved, unless otherwise indicated.