Neutrality and epistasis in program space

Joseph Renzullo, Westley Weimer, Melanie Moses, Stephanie Forrest

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Citations (Scopus)

Abstract

Neutral networks in biology often contain diverse solutions with equal fitness, which can be useful when environments (requirements) change over time. In this paper, we present a method for studying neutral networks in software. In these networks, we find multiple solutions to held-out test cases (latent bugs), suggesting that neutral software networks also exhibit relevant diversity. We also observe instances of positive epistasis between random mutations, i.e. interactions that collectively increase fitness. Positive epistasis is rare as a fraction of the total search space but significant as a fraction of the objective space: 9% of the repairs we found to look (and 4.63% across all programs analyzed) were produced by positive interactions between mutations. Further, the majority (62.50%) of unique repairs are instances of positive epistasis.

Original languageEnglish (US)
Title of host publicationProceedings - 2018 ACM/IEEE 4th International Genetic Improvement Workshop, GI 2018
PublisherIEEE Computer Society
Pages1-8
Number of pages8
VolumePart F139756
ISBN (Print)9781450357531
DOIs
StatePublished - Jun 2 2018
Event2018 ACM/IEEE 4th International Genetic Improvement Workshop, GI 2018, Held at 40th International Conference on Software Engineering, ICSE 2018 - Gothenburg, Sweden
Duration: Jun 2 2018 → …

Other

Other2018 ACM/IEEE 4th International Genetic Improvement Workshop, GI 2018, Held at 40th International Conference on Software Engineering, ICSE 2018
CountrySweden
CityGothenburg
Period6/2/18 → …

Fingerprint

Repair

Keywords

  • automated software engineering
  • biological networks
  • network science
  • software evolution
  • software testing and debugging

ASJC Scopus subject areas

  • Software

Cite this

Renzullo, J., Weimer, W., Moses, M., & Forrest, S. (2018). Neutrality and epistasis in program space. In Proceedings - 2018 ACM/IEEE 4th International Genetic Improvement Workshop, GI 2018 (Vol. Part F139756, pp. 1-8). IEEE Computer Society. https://doi.org/10.1145/3194810.3194812

Neutrality and epistasis in program space. / Renzullo, Joseph; Weimer, Westley; Moses, Melanie; Forrest, Stephanie.

Proceedings - 2018 ACM/IEEE 4th International Genetic Improvement Workshop, GI 2018. Vol. Part F139756 IEEE Computer Society, 2018. p. 1-8.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Renzullo, J, Weimer, W, Moses, M & Forrest, S 2018, Neutrality and epistasis in program space. in Proceedings - 2018 ACM/IEEE 4th International Genetic Improvement Workshop, GI 2018. vol. Part F139756, IEEE Computer Society, pp. 1-8, 2018 ACM/IEEE 4th International Genetic Improvement Workshop, GI 2018, Held at 40th International Conference on Software Engineering, ICSE 2018, Gothenburg, Sweden, 6/2/18. https://doi.org/10.1145/3194810.3194812
Renzullo J, Weimer W, Moses M, Forrest S. Neutrality and epistasis in program space. In Proceedings - 2018 ACM/IEEE 4th International Genetic Improvement Workshop, GI 2018. Vol. Part F139756. IEEE Computer Society. 2018. p. 1-8 https://doi.org/10.1145/3194810.3194812
Renzullo, Joseph ; Weimer, Westley ; Moses, Melanie ; Forrest, Stephanie. / Neutrality and epistasis in program space. Proceedings - 2018 ACM/IEEE 4th International Genetic Improvement Workshop, GI 2018. Vol. Part F139756 IEEE Computer Society, 2018. pp. 1-8
@inproceedings{aac4515e2c4f4d6a81bb4aefa4a739c8,
title = "Neutrality and epistasis in program space",
abstract = "Neutral networks in biology often contain diverse solutions with equal fitness, which can be useful when environments (requirements) change over time. In this paper, we present a method for studying neutral networks in software. In these networks, we find multiple solutions to held-out test cases (latent bugs), suggesting that neutral software networks also exhibit relevant diversity. We also observe instances of positive epistasis between random mutations, i.e. interactions that collectively increase fitness. Positive epistasis is rare as a fraction of the total search space but significant as a fraction of the objective space: 9{\%} of the repairs we found to look (and 4.63{\%} across all programs analyzed) were produced by positive interactions between mutations. Further, the majority (62.50{\%}) of unique repairs are instances of positive epistasis.",
keywords = "automated software engineering, biological networks, network science, software evolution, software testing and debugging",
author = "Joseph Renzullo and Westley Weimer and Melanie Moses and Stephanie Forrest",
year = "2018",
month = "6",
day = "2",
doi = "10.1145/3194810.3194812",
language = "English (US)",
isbn = "9781450357531",
volume = "Part F139756",
pages = "1--8",
booktitle = "Proceedings - 2018 ACM/IEEE 4th International Genetic Improvement Workshop, GI 2018",
publisher = "IEEE Computer Society",

}

TY - GEN

T1 - Neutrality and epistasis in program space

AU - Renzullo, Joseph

AU - Weimer, Westley

AU - Moses, Melanie

AU - Forrest, Stephanie

PY - 2018/6/2

Y1 - 2018/6/2

N2 - Neutral networks in biology often contain diverse solutions with equal fitness, which can be useful when environments (requirements) change over time. In this paper, we present a method for studying neutral networks in software. In these networks, we find multiple solutions to held-out test cases (latent bugs), suggesting that neutral software networks also exhibit relevant diversity. We also observe instances of positive epistasis between random mutations, i.e. interactions that collectively increase fitness. Positive epistasis is rare as a fraction of the total search space but significant as a fraction of the objective space: 9% of the repairs we found to look (and 4.63% across all programs analyzed) were produced by positive interactions between mutations. Further, the majority (62.50%) of unique repairs are instances of positive epistasis.

AB - Neutral networks in biology often contain diverse solutions with equal fitness, which can be useful when environments (requirements) change over time. In this paper, we present a method for studying neutral networks in software. In these networks, we find multiple solutions to held-out test cases (latent bugs), suggesting that neutral software networks also exhibit relevant diversity. We also observe instances of positive epistasis between random mutations, i.e. interactions that collectively increase fitness. Positive epistasis is rare as a fraction of the total search space but significant as a fraction of the objective space: 9% of the repairs we found to look (and 4.63% across all programs analyzed) were produced by positive interactions between mutations. Further, the majority (62.50%) of unique repairs are instances of positive epistasis.

KW - automated software engineering

KW - biological networks

KW - network science

KW - software evolution

KW - software testing and debugging

UR - http://www.scopus.com/inward/record.url?scp=85054619081&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85054619081&partnerID=8YFLogxK

U2 - 10.1145/3194810.3194812

DO - 10.1145/3194810.3194812

M3 - Conference contribution

AN - SCOPUS:85054619081

SN - 9781450357531

VL - Part F139756

SP - 1

EP - 8

BT - Proceedings - 2018 ACM/IEEE 4th International Genetic Improvement Workshop, GI 2018

PB - IEEE Computer Society

ER -