Faster Convergence with Lexicase Selection in Tree-Based Automated Machine Learning

Nicholas Matsumoto; Anil Kumar Saini; Pedro Ribeiro; Hyunjun Choi; Alena Orlenko; Leo Pekka Lyytikäinen; Jari O. Laurikka; Terho Lehtimäki; Sandra Batista; Jason H. Moore

doi:10.1007/978-3-031-29573-7_11

Faster Convergence with Lexicase Selection in Tree-Based Automated Machine Learning

Nicholas Matsumoto, Anil Kumar Saini, Pedro Ribeiro, Hyunjun Choi, Alena Orlenko, Leo Pekka Lyytikäinen, Jari O. Laurikka, Terho Lehtimäki, Sandra Batista, Jason H. Moore

TAYS Sydänkeskus Oy VA

Tutkimustuotos: Konferenssiartikkeli › Scientific › vertaisarvioitu

Abstrakti

In many evolutionary computation systems, parent selection methods can affect, among other things, convergence to a solution. In this paper, we present a study comparing the role of two commonly used parent selection methods in evolving machine learning pipelines in an automated machine learning system called Tree-based Pipeline Optimization Tool (TPOT). Specifically, we demonstrate, using experiments on multiple datasets, that lexicase selection leads to significantly faster convergence as compared to NSGA-II in TPOT. We also compare the exploration of parts of the search space by these selection methods using a trie data structure that contains information about the pipelines explored in a particular run.

Alkuperäiskieli	Englanti
Otsikko	Genetic Programming
Alaotsikko	26th European Conference, EuroGP 2023, Held as Part of EvoStar 2023, Proceedings
Toimittajat	Gisele Pappa, Mario Giacobini, Zdenek Vasicek
Kustantaja	Springer
Sivut	165-181
Sivumäärä	17
ISBN (elektroninen)	978-3-031-29573-7
ISBN (painettu)	978-3-031-29572-0
DOI - pysyväislinkit	https://doi.org/10.1007/978-3-031-29573-7_11
Tila	Julkaistu - 2023
OKM-julkaisutyyppi	A4 Artikkeli konferenssijulkaisussa
Tapahtuma	European Conference on Genetic Programming (Part of EvoStar) - Brno, Tshekki Kesto: 12 huhtik. 2023 → 14 huhtik. 2023 Konferenssinumero: 6th

Julkaisusarja

Nimi	Lecture Notes in Computer Science
Vuosikerta	13986
ISSN (painettu)	0302-9743
ISSN (elektroninen)	1611-3349

Conference

Conference	European Conference on Genetic Programming (Part of EvoStar)
Lyhennettä	EuroGP
Maa/Alue	Tshekki
Kaupunki	Brno
Ajanjakso	12/04/23 → 14/04/23

Julkaisufoorumi-taso

Jufo-taso 1

!!ASJC Scopus subject areas

Theoretical Computer Science
Yleinen tietojenkäsittelytiede

Pääsy asiakirjaan

10.1007/978-3-031-29573-7_11

Siteeraa tätä

Matsumoto, N., Saini, A. K., Ribeiro, P., Choi, H., Orlenko, A., Lyytikäinen, L. P., Laurikka, J. O., Lehtimäki, T., Batista, S., & Moore, J. H. (2023). Faster Convergence with Lexicase Selection in Tree-Based Automated Machine Learning. teoksessa G. Pappa, M. Giacobini, & Z. Vasicek (Toimittajat), Genetic Programming: 26th European Conference, EuroGP 2023, Held as Part of EvoStar 2023, Proceedings (Sivut 165-181). (Lecture Notes in Computer Science; Vuosikerta 13986). Springer. https://doi.org/10.1007/978-3-031-29573-7_11

Matsumoto, Nicholas ; Saini, Anil Kumar ; Ribeiro, Pedro et al. / Faster Convergence with Lexicase Selection in Tree-Based Automated Machine Learning. Genetic Programming: 26th European Conference, EuroGP 2023, Held as Part of EvoStar 2023, Proceedings. Toimittaja / Gisele Pappa ; Mario Giacobini ; Zdenek Vasicek. Springer, 2023. Sivut 165-181 (Lecture Notes in Computer Science).

@inproceedings{889f61a1d83043f38eb595f663edbbe6,

title = "Faster Convergence with Lexicase Selection in Tree-Based Automated Machine Learning",

abstract = "In many evolutionary computation systems, parent selection methods can affect, among other things, convergence to a solution. In this paper, we present a study comparing the role of two commonly used parent selection methods in evolving machine learning pipelines in an automated machine learning system called Tree-based Pipeline Optimization Tool (TPOT). Specifically, we demonstrate, using experiments on multiple datasets, that lexicase selection leads to significantly faster convergence as compared to NSGA-II in TPOT. We also compare the exploration of parts of the search space by these selection methods using a trie data structure that contains information about the pipelines explored in a particular run.",

keywords = "Convergence, Lexicase, NSGA-II, Parent Selection, Trie",

author = "Nicholas Matsumoto and Saini, {Anil Kumar} and Pedro Ribeiro and Hyunjun Choi and Alena Orlenko and Lyytik{\"a}inen, {Leo Pekka} and Laurikka, {Jari O.} and Terho Lehtim{\"a}ki and Sandra Batista and Moore, {Jason H.}",

note = "Funding Information: Acknowledgements. This work is supported by National Institute of Health grants R01 LM010098 and R01 AG066833. Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.; European Conference on Genetic Programming (Part of EvoStar), EuroGP ; Conference date: 12-04-2023 Through 14-04-2023",

year = "2023",

doi = "10.1007/978-3-031-29573-7_11",

language = "English",

isbn = "978-3-031-29572-0",

series = "Lecture Notes in Computer Science",

publisher = "Springer",

pages = "165--181",

editor = "Gisele Pappa and Mario Giacobini and Zdenek Vasicek",

booktitle = "Genetic Programming",

}

Matsumoto, N, Saini, AK, Ribeiro, P, Choi, H, Orlenko, A, Lyytikäinen, LP , Laurikka, JO , Lehtimäki, T, Batista, S & Moore, JH 2023, Faster Convergence with Lexicase Selection in Tree-Based Automated Machine Learning. julkaisussa G Pappa, M Giacobini & Z Vasicek (toim), Genetic Programming: 26th European Conference, EuroGP 2023, Held as Part of EvoStar 2023, Proceedings. Lecture Notes in Computer Science, Vuosikerta 13986, Springer, Sivut 165-181, European Conference on Genetic Programming (Part of EvoStar), Brno, Tshekki, 12/04/23. https://doi.org/10.1007/978-3-031-29573-7_11

Faster Convergence with Lexicase Selection in Tree-Based Automated Machine Learning. / Matsumoto, Nicholas; Saini, Anil Kumar; Ribeiro, Pedro et al.
Genetic Programming: 26th European Conference, EuroGP 2023, Held as Part of EvoStar 2023, Proceedings. toim. / Gisele Pappa; Mario Giacobini; Zdenek Vasicek. Springer, 2023. s. 165-181 (Lecture Notes in Computer Science; Vuosikerta 13986).

Tutkimustuotos: Konferenssiartikkeli › Scientific › vertaisarvioitu

TY - GEN

T1 - Faster Convergence with Lexicase Selection in Tree-Based Automated Machine Learning

AU - Matsumoto, Nicholas

AU - Saini, Anil Kumar

AU - Ribeiro, Pedro

AU - Choi, Hyunjun

AU - Orlenko, Alena

AU - Lyytikäinen, Leo Pekka

AU - Laurikka, Jari O.

AU - Lehtimäki, Terho

AU - Batista, Sandra

AU - Moore, Jason H.

N1 - Conference code: 6th

PY - 2023

Y1 - 2023

N2 - In many evolutionary computation systems, parent selection methods can affect, among other things, convergence to a solution. In this paper, we present a study comparing the role of two commonly used parent selection methods in evolving machine learning pipelines in an automated machine learning system called Tree-based Pipeline Optimization Tool (TPOT). Specifically, we demonstrate, using experiments on multiple datasets, that lexicase selection leads to significantly faster convergence as compared to NSGA-II in TPOT. We also compare the exploration of parts of the search space by these selection methods using a trie data structure that contains information about the pipelines explored in a particular run.

AB - In many evolutionary computation systems, parent selection methods can affect, among other things, convergence to a solution. In this paper, we present a study comparing the role of two commonly used parent selection methods in evolving machine learning pipelines in an automated machine learning system called Tree-based Pipeline Optimization Tool (TPOT). Specifically, we demonstrate, using experiments on multiple datasets, that lexicase selection leads to significantly faster convergence as compared to NSGA-II in TPOT. We also compare the exploration of parts of the search space by these selection methods using a trie data structure that contains information about the pipelines explored in a particular run.

KW - Convergence

KW - Lexicase

KW - NSGA-II

KW - Parent Selection

KW - Trie

U2 - 10.1007/978-3-031-29573-7_11

DO - 10.1007/978-3-031-29573-7_11

M3 - Conference contribution

AN - SCOPUS:85152558753

SN - 978-3-031-29572-0

T3 - Lecture Notes in Computer Science

SP - 165

EP - 181

BT - Genetic Programming

A2 - Pappa, Gisele

A2 - Giacobini, Mario

A2 - Vasicek, Zdenek

PB - Springer

T2 - European Conference on Genetic Programming (Part of EvoStar)

Y2 - 12 April 2023 through 14 April 2023

ER -

Matsumoto N, Saini AK, Ribeiro P, Choi H, Orlenko A, Lyytikäinen LP et al. Faster Convergence with Lexicase Selection in Tree-Based Automated Machine Learning. julkaisussa Pappa G, Giacobini M, Vasicek Z, toimittajat, Genetic Programming: 26th European Conference, EuroGP 2023, Held as Part of EvoStar 2023, Proceedings. Springer. 2023. s. 165-181. (Lecture Notes in Computer Science). doi: 10.1007/978-3-031-29573-7_11