Home
Resource
Model-based Bayesian reinforcement learning with generalized priors
PDF
PDF format is widely accepted and good for printing.
Plug-in required
PDF-1(3.64 MB)
Citation & Export
View Usage Statistics
Staff View

Citation & Export
Hide

Simple citation

Asmuth, John Thomas. Model-based Bayesian reinforcement learning with generalized priors. Retrieved from https://doi.org/doi:10.7282/T3HX1B9M

Export

Click here for information about Citation Management Tools at Rutgers.

Statistics
Hide

Description

TitleModel-based Bayesian reinforcement learning with generalized priors
NameAsmuth, John Thomas (author); Littman, Michael L (chair); Pavlovich, Vladimir (internal member); Hirsh, Haym (internal member); Poupart, Pascal (outside member); Rutgers University; Graduate School - New Brunswick
Date Created2013
Other Date2013-05 (degree)
SubjectComputer Science, Reinforcement learning, Bayesian statistical decision theory
Extentxx, 161 p. : ill.
DescriptionEffectively leveraging model structure in reinforcement learning is a difficult task, but failure to do so can result in computer agents that repeatedly take sub-optimal actions, despite having enough information to perform better. The Bayesian approach is a principled and well-studied method for leveraging model structure, and it is useful to use in the reinforcement learning setting. This dissertation studies different methods for bringing the Bayesian approach to bear for model-based reinforcement learning agents, as well as different models that can be used. The contributions include several examples of models that can be used for learning MDPs, and two novel algorithms, and their analyses, for using those models for efficient exploration: BOSS and BFS3. The Bayesian approach to model-based reinforcement learning provides a principled method for incorporating prior knowledge into the design of an agent, and allows the designer to separate the problems of planning, learning and exploration. The BOSS and BFS3 algorithms are efficient (polynomial time) mechanisms for decision making within this framework with provable bounds on their accuracy.
NotePh.D.
NoteIncludes bibliographical references
NoteIncludes vita
Noteby John Thomas Asmuth
Genretheses, ETD doctoral
Persistent URLhttps://doi.org/doi:10.7282/T3HX1B9M
Languageeng
CollectionGraduate School - New Brunswick Electronic Theses and Dissertations
Organization NameRutgers, The State University of New Jersey
RightsThe author owns the copyright to this work.
Version 8.5.5
Rutgers University Libraries - Copyright ©2024