Influence-Based Abstraction in Deep Reinforcement Learning

Conference Paper (2019)
Author(s)

Miguel Suau de Castro (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Elena Congeduti (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Rolf Starre (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Aleksander Czechowski (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Frans Oliehoek (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Research Group
Interactive Intelligence
URL related publication
https://ala2019.vub.ac.be/
More Info
expand_more
Publication Year
2019
Language
English
Research Group
Interactive Intelligence
Bibliographical Note
Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.
Event
ALA 2019 - Workshop at AAMAS 2019 (2019-05-13 - 2019-05-14), Montreal, Canada
Downloads counter
297
Collections
Institutional Repository
Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Abstract

thousands, or even millions of state variables. Unfortunately, applying reinforcement learning algorithms to handle complex tasks becomes more and more challenging as the number of state variables increases. In this paper, we build on the concept of influence-based abstraction which tries to tackle such scalability issues by decomposing large systems into small regions. We explore this method in the context of deep reinforcement learning, showing that by keeping track of a small set of variables in the history of previous actions and observations we can learn policies that can effectively control a local region in the global system.

Files

Suau19ALA.pdf
(pdf | 1.03 Mb)
- Embargo expired in 01-05-2020
Unspecified