Re-understanding Finite-State Representations of Recurrent Policy Networks

Graduate Thesis Or Dissertation

Public Deposited

Citeable URL: https://ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/1r66j8037

Descriptions

Attribute Name	Values
Creator	Danesh, Mohamad H.
Abstract	We propose an approach for understanding control policies represented as recurrent neural networks. Recent work has approached this problem by transforming such recurrent policy networks into finite-state machines (FSM) and then analyzing the equivalent minimized FSM. While this led to interesting insights, the minimization process can obscure a deeper understanding of a machine's operation by merging states that are semantically distinct. To address this issue, we introduce an analysis approach that starts with an unminimized FSM and applies more-interpretable reductions that preserve the key decision points of the policy. We also contribute an attention tool to attain a deeper understanding of the role of observations in the decisions. Our case studies on 7 Atari games and 3 control benchmarks demonstrate that the approach can reveal insights that have not been previously noticed.
Contributor	Koul, Anurag Fern, Alan P. Khorram, Saeed
License	All rights reserved
Resource Type	Masters Thesis
Date Issued	2021-06-09
Degree Level	Master's
Degree Name	Master of Science (M.S.)
Degree Field	Computer Science
Degree Grantor	Oregon State University
Commencement Year	2019
Advisor	Fern, Alan P.
Committee Member	Wong, Weng-Keen Tadepalli, Prasad Stefan, Lee
Academic Affiliation	Electrical Engineering and Computer Science
Rights Statement	In Copyright
Publisher	Oregon State University
Peer Reviewed	No
Language	English [eng]

Parents:

This work has no parents.

In Collection:

Thumbnail	Title	Date Uploaded	Visibility	Actions
	DaneshMohamadHosein2021.pdf	2021-06-10	Public	Download