Asymptotically optimal agents

Tor Lattimore; Marcus Hutter

doi:10.1007/978-3-642-24412-4_29

Asymptotically optimal agents

Tor Lattimore^*, Marcus Hutter

^*Corresponding author for this work

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

23 Citations (Scopus)

Abstract

Artificial general intelligence aims to create agents capable of learning to solve arbitrary interesting problems. We define two versions of asymptotic optimality and prove that no agent can satisfy the strong version while in some cases, depending on discounting, there does exist a non-computable weak asymptotically optimal agent.

Original language	English
Title of host publication	Algorithmic Learning Theory - 22nd International Conference, ALT 2011, Proceedings
Pages	368-382
Number of pages	15
DOIs	https://doi.org/10.1007/978-3-642-24412-4_29
Publication status	Published - 2011
Event	22nd International Conference on Algorithmic Learning Theory, ALT 2011 - Espoo, Finland Duration: 5 Oct 2011 → 7 Oct 2011

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	6925 LNAI
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	22nd International Conference on Algorithmic Learning Theory, ALT 2011
Country/Territory	Finland
City	Espoo
Period	5/10/11 → 7/10/11

Access to Document

10.1007/978-3-642-24412-4_29

Cite this

@inproceedings{3d5eac56afac42b8828db070723a39a9,

title = "Asymptotically optimal agents",

abstract = "Artificial general intelligence aims to create agents capable of learning to solve arbitrary interesting problems. We define two versions of asymptotic optimality and prove that no agent can satisfy the strong version while in some cases, depending on discounting, there does exist a non-computable weak asymptotically optimal agent.",

keywords = "Rational agents, artificial general intelligence, asymptotic optimality, general discounting, reinforcement learning, sequential decision theory",

author = "Tor Lattimore and Marcus Hutter",

year = "2011",

doi = "10.1007/978-3-642-24412-4_29",

language = "English",

isbn = "9783642244117",

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

pages = "368--382",

booktitle = "Algorithmic Learning Theory - 22nd International Conference, ALT 2011, Proceedings",

note = "22nd International Conference on Algorithmic Learning Theory, ALT 2011 ; Conference date: 05-10-2011 Through 07-10-2011",

}

Lattimore, T & Hutter, M 2011, Asymptotically optimal agents. in Algorithmic Learning Theory - 22nd International Conference, ALT 2011, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 6925 LNAI, pp. 368-382, 22nd International Conference on Algorithmic Learning Theory, ALT 2011, Espoo, Finland, 5/10/11. https://doi.org/10.1007/978-3-642-24412-4_29

Asymptotically optimal agents. / Lattimore, Tor; Hutter, Marcus.
Algorithmic Learning Theory - 22nd International Conference, ALT 2011, Proceedings. 2011. p. 368-382 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6925 LNAI).

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

TY - GEN

T1 - Asymptotically optimal agents

AU - Lattimore, Tor

AU - Hutter, Marcus

PY - 2011

Y1 - 2011

N2 - Artificial general intelligence aims to create agents capable of learning to solve arbitrary interesting problems. We define two versions of asymptotic optimality and prove that no agent can satisfy the strong version while in some cases, depending on discounting, there does exist a non-computable weak asymptotically optimal agent.

AB - Artificial general intelligence aims to create agents capable of learning to solve arbitrary interesting problems. We define two versions of asymptotic optimality and prove that no agent can satisfy the strong version while in some cases, depending on discounting, there does exist a non-computable weak asymptotically optimal agent.

KW - Rational agents

KW - artificial general intelligence

KW - asymptotic optimality

KW - general discounting

KW - reinforcement learning

KW - sequential decision theory

UR - http://www.scopus.com/inward/record.url?scp=80054111284&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-24412-4_29

DO - 10.1007/978-3-642-24412-4_29

M3 - Conference contribution

SN - 9783642244117

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

SP - 368

EP - 382

BT - Algorithmic Learning Theory - 22nd International Conference, ALT 2011, Proceedings

T2 - 22nd International Conference on Algorithmic Learning Theory, ALT 2011

Y2 - 5 October 2011 through 7 October 2011

ER -

Asymptotically optimal agents

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this