Market-based reinforcement learning in partially observable worlds

Ivo Kwee; Marcus Hutter; Jürgen Schmidhuber

doi:10.1007/3-540-44668-0_120

Market-based reinforcement learning in partially observable worlds

Ivo Kwee, Marcus Hutter, Jürgen Schmidhuber

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

11 Citations (Scopus)

Abstract

Unlike traditional reinforcement learning (RL), marketbased RL is in principle applicable to worlds described by partially observable Markov Decision Processes (POMDPs), where an agent needs to learn short-term memories of relevant previous events in order to execute optimal actions. Most previous work, however, has focused on reactive settings (MDPs) instead of POMDPs. Here we reimplement a recent approach to market-based RL and for the first time evaluate it in a toy POMDP setting.

Original language	English
Title of host publication	Artificial Neural Networks - ICANN 2001 - International Conference, Proceedings
Editors	Kurt Hornik, Georg Dorffner, Horst Bischof
Publisher	Springer Verlag
Pages	865-873
Number of pages	9
ISBN (Print)	3540424865, 9783540446682
DOIs	https://doi.org/10.1007/3-540-44668-0_120
Publication status	Published - 2001
Externally published	Yes
Event	International Conference on Artificial Neural Networks, ICANN 2001 - Vienna, Austria Duration: 21 Aug 2001 → 25 Aug 2001

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	2130
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	International Conference on Artificial Neural Networks, ICANN 2001
Country/Territory	Austria
City	Vienna
Period	21/08/01 → 25/08/01

Access to Document

10.1007/3-540-44668-0_120

Cite this

Kwee, I., Hutter, M., & Schmidhuber, J. (2001). Market-based reinforcement learning in partially observable worlds. In K. Hornik, G. Dorffner, & H. Bischof (Eds.), Artificial Neural Networks - ICANN 2001 - International Conference, Proceedings (pp. 865-873). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 2130). Springer Verlag. https://doi.org/10.1007/3-540-44668-0_120

Kwee, Ivo ; Hutter, Marcus ; Schmidhuber, Jürgen. / Market-based reinforcement learning in partially observable worlds. Artificial Neural Networks - ICANN 2001 - International Conference, Proceedings. editor / Kurt Hornik ; Georg Dorffner ; Horst Bischof. Springer Verlag, 2001. pp. 865-873 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{cc3e0f4c6cb745e0bd2e4031e5bd43c8,

title = "Market-based reinforcement learning in partially observable worlds",

abstract = "Unlike traditional reinforcement learning (RL), marketbased RL is in principle applicable to worlds described by partially observable Markov Decision Processes (POMDPs), where an agent needs to learn short-term memories of relevant previous events in order to execute optimal actions. Most previous work, however, has focused on reactive settings (MDPs) instead of POMDPs. Here we reimplement a recent approach to market-based RL and for the first time evaluate it in a toy POMDP setting.",

author = "Ivo Kwee and Marcus Hutter and J{\"u}rgen Schmidhuber",

note = "Publisher Copyright: {\textcopyright} Springer-Verlag Berlin Heidelberg 2001.; International Conference on Artificial Neural Networks, ICANN 2001 ; Conference date: 21-08-2001 Through 25-08-2001",

year = "2001",

doi = "10.1007/3-540-44668-0_120",

language = "English",

isbn = "3540424865",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "865--873",

editor = "Kurt Hornik and Georg Dorffner and Horst Bischof",

booktitle = "Artificial Neural Networks - ICANN 2001 - International Conference, Proceedings",

address = "Germany",

}

Kwee, I, Hutter, M & Schmidhuber, J 2001, Market-based reinforcement learning in partially observable worlds. in K Hornik, G Dorffner & H Bischof (eds), Artificial Neural Networks - ICANN 2001 - International Conference, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 2130, Springer Verlag, pp. 865-873, International Conference on Artificial Neural Networks, ICANN 2001, Vienna, Austria, 21/08/01. https://doi.org/10.1007/3-540-44668-0_120

Market-based reinforcement learning in partially observable worlds. / Kwee, Ivo; Hutter, Marcus; Schmidhuber, Jürgen.
Artificial Neural Networks - ICANN 2001 - International Conference, Proceedings. ed. / Kurt Hornik; Georg Dorffner; Horst Bischof. Springer Verlag, 2001. p. 865-873 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 2130).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Market-based reinforcement learning in partially observable worlds

AU - Kwee, Ivo

AU - Hutter, Marcus

AU - Schmidhuber, Jürgen

N1 - Publisher Copyright: © Springer-Verlag Berlin Heidelberg 2001.

PY - 2001

Y1 - 2001

N2 - Unlike traditional reinforcement learning (RL), marketbased RL is in principle applicable to worlds described by partially observable Markov Decision Processes (POMDPs), where an agent needs to learn short-term memories of relevant previous events in order to execute optimal actions. Most previous work, however, has focused on reactive settings (MDPs) instead of POMDPs. Here we reimplement a recent approach to market-based RL and for the first time evaluate it in a toy POMDP setting.

AB - Unlike traditional reinforcement learning (RL), marketbased RL is in principle applicable to worlds described by partially observable Markov Decision Processes (POMDPs), where an agent needs to learn short-term memories of relevant previous events in order to execute optimal actions. Most previous work, however, has focused on reactive settings (MDPs) instead of POMDPs. Here we reimplement a recent approach to market-based RL and for the first time evaluate it in a toy POMDP setting.

UR - http://www.scopus.com/inward/record.url?scp=84865410615&partnerID=8YFLogxK

U2 - 10.1007/3-540-44668-0_120

DO - 10.1007/3-540-44668-0_120

M3 - Conference contribution

SN - 3540424865

SN - 9783540446682

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 865

EP - 873

BT - Artificial Neural Networks - ICANN 2001 - International Conference, Proceedings

A2 - Hornik, Kurt

A2 - Dorffner, Georg

A2 - Bischof, Horst

PB - Springer Verlag

T2 - International Conference on Artificial Neural Networks, ICANN 2001

Y2 - 21 August 2001 through 25 August 2001

ER -

Kwee I, Hutter M, Schmidhuber J. Market-based reinforcement learning in partially observable worlds. In Hornik K, Dorffner G, Bischof H, editors, Artificial Neural Networks - ICANN 2001 - International Conference, Proceedings. Springer Verlag. 2001. p. 865-873. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/3-540-44668-0_120

Market-based reinforcement learning in partially observable worlds

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this