Deriving Subgoals Autonomously to Accelerate Learning in Sparse Reward Domains.

Dann, M, Zambetta, F and Thangarajah, J 2019, 'Deriving Subgoals Autonomously to Accelerate Learning in Sparse Reward Domains.', in Proceedings of the Thirty-Third AAAI Conference on Artificial Intelligence (AAAI 2019), Hawaii, United States, 27 January - 1 February 2019, pp. 881-889.


Document type: Conference Paper
Collection: Conference Papers

Title Deriving Subgoals Autonomously to Accelerate Learning in Sparse Reward Domains.
Author(s) Dann, M
Zambetta, F
Thangarajah, J
Year 2019
Conference name AAAI 2019
Conference location Hawaii, United States
Conference dates 27 January - 1 February 2019
Proceedings title Proceedings of the Thirty-Third AAAI Conference on Artificial Intelligence (AAAI 2019)
Publisher AAAI Press
Place of publication California, United States
Start page 881
End page 889
Total pages 9
Abstract Sparse reward games, such as the infamous Montezumas Revenge, pose a significant challenge for Reinforcement Learning (RL) agents. Hierarchical RL, which promotes efficient exploration via subgoals, has shown promise in these games. However, existing agents rely either on human domain knowledge or slow autonomous methods to derive suitable subgoals. In this work, we describe a new, autonomous approach for deriving subgoals from raw pixels that is more efficient than competing methods. We propose a novel intrinsic reward scheme for exploiting the derived subgoals, applying it to three Atari games with sparse rewards. Our agents performance is comparable to that of state-of-the-art methods, demonstrating the usefulness of the subgoals found.
Subjects Adaptive Agents and Intelligent Robotics
DOI - identifier 10.1609/aaai.v33i01.3301881
Copyright notice Copyright © 2019, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.
ISBN 9781577358091
Versions
Version Filter Type
Citation counts: TR Web of Science Citation Count  Cited 0 times in Thomson Reuters Web of Science Article
Altmetric details:
Access Statistics: 23 Abstract Views  -  Detailed Statistics
Created: Wed, 23 Oct 2019, 08:59:00 EST by Catalyst Administrator
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us