Measuring and optimizing behavioral complexity for evolutionary reinforcement learning

Faustino J. Gomez; Julian Togelius; Juergen Schmidhuber

doi:10.1007/978-3-642-04277-5_77

Measuring and optimizing behavioral complexity for evolutionary reinforcement learning

Faustino J. Gomez, Julian Togelius, Juergen Schmidhuber

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

9 Scopus citations

Abstract

Model complexity is key concern to any artificial learning system due its critical impact on generalization. However, EC research has only focused phenotype structural complexity for static problems. For sequential decision tasks, phenotypes that are very similar in structure, can produce radically different behaviors, and the trade-off between fitness and complexity in this context is not clear. In this paper, behavioral complexity is measured explicitly using compression, and used as a separate objective to be optimized (not as an additional regularization term in a scalar fitness), in order to study this trade-off directly. © 2009 Springer Berlin Heidelberg.

Original language	English (US)
Title of host publication	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages	765-774
Number of pages	10
DOIs	https://doi.org/10.1007/978-3-642-04277-5_77
State	Published - Nov 30 2009
Externally published	Yes

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/978-3-642-04277-5_77

Cite this

@inproceedings{b02fd32ce7394bb5801779992602b565,

title = "Measuring and optimizing behavioral complexity for evolutionary reinforcement learning",

abstract = "Model complexity is key concern to any artificial learning system due its critical impact on generalization. However, EC research has only focused phenotype structural complexity for static problems. For sequential decision tasks, phenotypes that are very similar in structure, can produce radically different behaviors, and the trade-off between fitness and complexity in this context is not clear. In this paper, behavioral complexity is measured explicitly using compression, and used as a separate objective to be optimized (not as an additional regularization term in a scalar fitness), in order to study this trade-off directly. {\textcopyright} 2009 Springer Berlin Heidelberg.",

author = "Gomez, {Faustino J.} and Julian Togelius and Juergen Schmidhuber",

note = "Generated from Scopus record by KAUST IRTS on 2022-09-14",

year = "2009",

month = nov,

day = "30",

doi = "10.1007/978-3-642-04277-5_77",

language = "English (US)",

isbn = "3642042767",

pages = "765--774",

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

}

Measuring and optimizing behavioral complexity for evolutionary reinforcement learning. / Gomez, Faustino J.; Togelius, Julian; Schmidhuber, Juergen.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 2009. p. 765-774.

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

TY - GEN

T1 - Measuring and optimizing behavioral complexity for evolutionary reinforcement learning

AU - Gomez, Faustino J.

AU - Togelius, Julian

AU - Schmidhuber, Juergen

N1 - Generated from Scopus record by KAUST IRTS on 2022-09-14

PY - 2009/11/30

Y1 - 2009/11/30

N2 - Model complexity is key concern to any artificial learning system due its critical impact on generalization. However, EC research has only focused phenotype structural complexity for static problems. For sequential decision tasks, phenotypes that are very similar in structure, can produce radically different behaviors, and the trade-off between fitness and complexity in this context is not clear. In this paper, behavioral complexity is measured explicitly using compression, and used as a separate objective to be optimized (not as an additional regularization term in a scalar fitness), in order to study this trade-off directly. © 2009 Springer Berlin Heidelberg.

AB - Model complexity is key concern to any artificial learning system due its critical impact on generalization. However, EC research has only focused phenotype structural complexity for static problems. For sequential decision tasks, phenotypes that are very similar in structure, can produce radically different behaviors, and the trade-off between fitness and complexity in this context is not clear. In this paper, behavioral complexity is measured explicitly using compression, and used as a separate objective to be optimized (not as an additional regularization term in a scalar fitness), in order to study this trade-off directly. © 2009 Springer Berlin Heidelberg.

UR - http://link.springer.com/10.1007/978-3-642-04277-5_77

UR - http://www.scopus.com/inward/record.url?scp=70450227694&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-04277-5_77

DO - 10.1007/978-3-642-04277-5_77

M3 - Conference contribution

SN - 3642042767

SP - 765

EP - 774

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

ER -

Measuring and optimizing behavioral complexity for evolutionary reinforcement learning

Abstract

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this