Please use this identifier to cite or link to this item:
https://hdl.handle.net/2440/123724
Citations | ||
Scopus | Web of Science® | Altmetric |
---|---|---|
?
|
?
|
Type: | Conference paper |
Title: | Reinforcement learning with attention that works: a self-supervised approach |
Author: | Manchin, A. Abbasnejad, E. Van Den Hengel, A. |
Citation: | Communications in Computer and Information Science, 2019 / Gedeon, T., Wong, K.W., Lee, M. (ed./s), vol.1143 CCIS, pp.223-230 |
Publisher: | Springer |
Publisher Place: | Switzerland |
Issue Date: | 2019 |
Series/Report no.: | Communications in Computer and Information Science; 1143 |
ISBN: | 9783030368012 |
ISSN: | 1865-0929 1865-0937 |
Conference Name: | International Conference on Neural Information Processing (ICONIP) (12 Dec 2019 - 15 Dec 2019 : Sydney, Australia) |
Editor: | Gedeon, T. Wong, K.W. Lee, M. |
Statement of Responsibility: | Anthony Manchin, Ehsan Abbasnejad, and Anton van den Hengel |
Abstract: | Attention models have had a significant positive impact on deep learning across a range of tasks. However previous attempts at integrating attention with reinforcement learning have failed to produce significant improvements. Unlike the selective attention models used in previous attempts, which constrain the attention via preconceived notions of importance, our implementation utilises the Markovian properties inherent in the state input. We propose the first combination of self attention and reinforcement learning that is capable of producing significant improvements, including new state of the art results in the Arcade Learning Environment. |
Keywords: | Reinforcement learning; Attention; Deep learning |
Rights: | © Springer Nature Switzerland AG 2019 |
DOI: | 10.1007/978-3-030-36802-9_25 |
Published version: | http://dx.doi.org/10.1007/978-3-030-36802-9_25 |
Appears in Collections: | Aurora harvest 4 Australian Institute for Machine Learning publications |
Files in This Item:
There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.