Please use this identifier to cite or link to this item:
Title: Competitive reinforcement learning in Atari games
Contributor(s): McKenzie, Mark (author); Loxley, Peter  (author)orcid ; Billingsley, William  (author)orcid ; Wong, Sebastien (author)
Publication Date: 2017
DOI: 10.1007/978-3-319-63004-5_2
Handle Link:
Abstract: This research describes a study into the ability of a state of the art reinforcement learning algorithm to learn to perform multiple tasks. We demonstrate that the limitation of learning to performing two tasks can be mitigated with a competitive training method. We show that this approach results in improved generalization of the system when performing unforeseen tasks. The learning agent assessed is an altered version of the DeepMind deep Q–learner network (DQN), which has been demonstrated to outperform human players for a number of Atari 2600 games. The key findings of this paper is that there were significant degradations in performance when learning more than one game, and how this varies depends on both similarity and the comparative complexity of the two games.
Publication Type: Conference Publication
Conference Name: 30th Australasian Joint Conference on Artificial Intelligence, AI 2017, Melbourne, VIC, AUS, 19-08-2017
Source of Publication: AI 2017: Advances in Artificial Intelligence, 10400(LNAI), p. 14-26
Publisher: Springer
Place of Publication: Germany
Field of Research (FOR): 080101 Adaptive Agents and Intelligent Robotics
080108 Neural, Evolutionary and Fuzzy Computation
Peer Reviewed: Yes
HERDC Category Description: E1 Refereed Scholarly Conference Publication
Series Name: Lecture Notes in Computer Science
Statistics to Oct 2018: Visitors: 26
Views: 43
Downloads: 0
Appears in Collections:Conference Publication

Files in This Item:
2 files
File Description SizeFormat 
Show full item record


checked on Nov 30, 2018

Page view(s)

checked on May 2, 2019
Google Media

Google ScholarTM



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