Home > Research > Publications & Outputs > MUMBO

Electronic data

  • 2006.12093v1

    Accepted author manuscript, 1.01 MB, PDF document

    Available under license: CC BY-NC: Creative Commons Attribution-NonCommercial 4.0 International License

Links

Text available via DOI:

Keywords

View graph of relations

MUMBO: MUlti-task Max-value Bayesian Optimization

Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSNConference contribution/Paperpeer-review

Published

Standard

MUMBO: MUlti-task Max-value Bayesian Optimization. / Moss, Henry B.; Leslie, David S.; Rayson, Paul.
Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2020, Proceedings. ed. / Frank Hutter; Kristian Kersting; Jefrey Lijffijt; Isabel Valera. Cham: Springer, 2020. p. 447-462 (Lecture Notes in Computer Science; Vol. 12459).

Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSNConference contribution/Paperpeer-review

Harvard

Moss, HB, Leslie, DS & Rayson, P 2020, MUMBO: MUlti-task Max-value Bayesian Optimization. in F Hutter, K Kersting, J Lijffijt & I Valera (eds), Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2020, Proceedings. Lecture Notes in Computer Science, vol. 12459, Springer, Cham, pp. 447-462. https://doi.org/10.1007/978-3-030-67664-3_27

APA

Moss, H. B., Leslie, D. S., & Rayson, P. (2020). MUMBO: MUlti-task Max-value Bayesian Optimization. In F. Hutter, K. Kersting, J. Lijffijt, & I. Valera (Eds.), Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2020, Proceedings (pp. 447-462). (Lecture Notes in Computer Science; Vol. 12459). Springer. https://doi.org/10.1007/978-3-030-67664-3_27

Vancouver

Moss HB, Leslie DS, Rayson P. MUMBO: MUlti-task Max-value Bayesian Optimization. In Hutter F, Kersting K, Lijffijt J, Valera I, editors, Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2020, Proceedings. Cham: Springer. 2020. p. 447-462. (Lecture Notes in Computer Science). doi: 10.1007/978-3-030-67664-3_27

Author

Moss, Henry B. ; Leslie, David S. ; Rayson, Paul. / MUMBO : MUlti-task Max-value Bayesian Optimization. Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2020, Proceedings. editor / Frank Hutter ; Kristian Kersting ; Jefrey Lijffijt ; Isabel Valera. Cham : Springer, 2020. pp. 447-462 (Lecture Notes in Computer Science).

Bibtex

@inproceedings{fe0f811514c84752948366521eb9fad0,
title = "MUMBO: MUlti-task Max-value Bayesian Optimization",
abstract = " We propose MUMBO, the first high-performing yet computationally efficient acquisition function for multi-task Bayesian optimization. Here, the challenge is to perform efficient optimization by evaluating low-cost functions somehow related to our true target function. This is a broad class of problems including the popular task of multi-fidelity optimization. However, while information-theoretic acquisition functions are known to provide state-of-the-art Bayesian optimization, existing implementations for multi-task scenarios have prohibitive computational requirements. Previous acquisition functions have therefore been suitable only for problems with both low-dimensional parameter spaces and function query costs sufficiently large to overshadow very significant optimization overheads. In this work, we derive a novel multi-task version of entropy search, delivering robust performance with low computational overheads across classic optimization challenges and multi-task hyper-parameter tuning. MUMBO is scalable and efficient, allowing multi-task Bayesian optimization to be deployed in problems with rich parameter and fidelity spaces. ",
keywords = "cs.LG, stat.ML",
author = "Moss, {Henry B.} and Leslie, {David S.} and Paul Rayson",
year = "2020",
month = sep,
day = "14",
doi = "10.1007/978-3-030-67664-3_27",
language = "English",
isbn = "9783030676636",
series = "Lecture Notes in Computer Science",
publisher = "Springer",
pages = "447--462",
editor = "Frank Hutter and Kristian Kersting and Jefrey Lijffijt and Isabel Valera",
booktitle = "Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2020, Proceedings",

}

RIS

TY - GEN

T1 - MUMBO

T2 - MUlti-task Max-value Bayesian Optimization

AU - Moss, Henry B.

AU - Leslie, David S.

AU - Rayson, Paul

PY - 2020/9/14

Y1 - 2020/9/14

N2 - We propose MUMBO, the first high-performing yet computationally efficient acquisition function for multi-task Bayesian optimization. Here, the challenge is to perform efficient optimization by evaluating low-cost functions somehow related to our true target function. This is a broad class of problems including the popular task of multi-fidelity optimization. However, while information-theoretic acquisition functions are known to provide state-of-the-art Bayesian optimization, existing implementations for multi-task scenarios have prohibitive computational requirements. Previous acquisition functions have therefore been suitable only for problems with both low-dimensional parameter spaces and function query costs sufficiently large to overshadow very significant optimization overheads. In this work, we derive a novel multi-task version of entropy search, delivering robust performance with low computational overheads across classic optimization challenges and multi-task hyper-parameter tuning. MUMBO is scalable and efficient, allowing multi-task Bayesian optimization to be deployed in problems with rich parameter and fidelity spaces.

AB - We propose MUMBO, the first high-performing yet computationally efficient acquisition function for multi-task Bayesian optimization. Here, the challenge is to perform efficient optimization by evaluating low-cost functions somehow related to our true target function. This is a broad class of problems including the popular task of multi-fidelity optimization. However, while information-theoretic acquisition functions are known to provide state-of-the-art Bayesian optimization, existing implementations for multi-task scenarios have prohibitive computational requirements. Previous acquisition functions have therefore been suitable only for problems with both low-dimensional parameter spaces and function query costs sufficiently large to overshadow very significant optimization overheads. In this work, we derive a novel multi-task version of entropy search, delivering robust performance with low computational overheads across classic optimization challenges and multi-task hyper-parameter tuning. MUMBO is scalable and efficient, allowing multi-task Bayesian optimization to be deployed in problems with rich parameter and fidelity spaces.

KW - cs.LG

KW - stat.ML

U2 - 10.1007/978-3-030-67664-3_27

DO - 10.1007/978-3-030-67664-3_27

M3 - Conference contribution/Paper

SN - 9783030676636

T3 - Lecture Notes in Computer Science

SP - 447

EP - 462

BT - Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2020, Proceedings

A2 - Hutter, Frank

A2 - Kersting, Kristian

A2 - Lijffijt, Jefrey

A2 - Valera, Isabel

PB - Springer

CY - Cham

ER -