Show simple item record

dc.contributor.advisorIsbell, Charles
dc.contributor.authorKumar, Saurabh
dc.date.accessioned2018-08-20T19:10:36Z
dc.date.available2018-08-20T19:10:36Z
dc.date.created2018-05
dc.date.submittedMay 2018
dc.identifier.urihttp://hdl.handle.net/1853/60339
dc.description.abstractTypical reinforcement learning (RL) agents learn to complete tasks specified by reward functions tailored to their domain. As such, the policies they learn do not generalize even to similar domains. To address this issue, we develop a framework through which a deep RL agent learns to generalize policies from smaller, simpler domains to more complex ones using a recurrent attention mechanism. The task is presented to the agent as an image and an instruction specifying the goal. This meta-controller guides the agent towards its goal by designing a sequence of smaller subtasks on the part of the state space within the attention, effectively decomposing it. As a baseline, we consider a setup without attention as well. Our experiments show that the meta-controller learns to create sub-goals within the attention. These results have implications for human-robot interactive applications, in which a robot can transfer skills it has learned in one task to another one and be robust to variability in its environment.
dc.format.mimetypeapplication/pdf
dc.language.isoen_US
dc.publisherGeorgia Institute of Technology
dc.subjectReinforcement learning
dc.subjectPolicy gradients
dc.subjectRecurrence
dc.subjectAttention mechanism
dc.titleState Space Decomposition in Reinforcement Learning
dc.typeUndergraduate Research Option Thesis
dc.description.degreeUndergraduate
dc.contributor.departmentComputer Science
thesis.degree.levelUndergraduate
dc.contributor.committeeMemberRiedl, Mark
dc.date.updated2018-08-20T19:10:36Z


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record