An Asynchronous Implementation of the Limited Memory CMA-ES

Viktor Arkhipov; Maxim Buzdalov; A. A. Shalyto

doi:10.1109/ICMLA.2015.97

An Asynchronous Implementation of the Limited Memory CMA-ES

Viktor Arkhipov, Maxim Buzdalov, A. A. Shalyto

ITMO University

Research output: Chapter in Book/Report/Conference proceeding › Conference Proceeding (Non-Journal item)

Abstract

We present our asynchronous implementation of the LM-CMA-ES algorithm, which is a modern evolution strategy for solving complex large-scale continuous optimization problems. Our implementation brings the best results when the number of cores is relatively high and the computational complexity of the fitness function is also high. The experiments with benchmark functions show that it is able to overcome its origin on the Sphere function, reaches certain thresholds faster on the Rosenbrock and Ellipsoid function, and surprisingly performs much better than the original version on the Rastrigin function.

Original language	English
Title of host publication	2015 IEEE 14th International Conference on Machine Learning and Applications (ICMLA)
Publisher	IEEE Press
Pages	707-712
Number of pages	6
ISBN (Electronic)	978-1-5090-0287-0, 1509002871
DOIs	https://doi.org/10.1109/ICMLA.2015.97
Publication status	Published - 03 Mar 2016
Externally published	Yes
Event	IEEE 14th International Conference on Machine Learning and Applications (ICMLA) - Miami, United States of America Duration: 09 Dec 2015 → 11 Dec 2015

Conference

Conference	IEEE 14th International Conference on Machine Learning and Applications (ICMLA)
Country/Territory	United States of America
City	Miami
Period	09 Dec 2015 → 11 Dec 2015

Keywords

CMA-ES
asynchronous algorithms
parallel algorithms

Access to Document

10.1109/ICMLA.2015.97

http://arxiv.org/abs/1510.00419

Permanent link

Persistent link

Cite this

@inproceedings{551378e1e30842cc83a953cc7253c2e2,

title = "An Asynchronous Implementation of the Limited Memory CMA-ES",

abstract = "We present our asynchronous implementation of the LM-CMA-ES algorithm, which is a modern evolution strategy for solving complex large-scale continuous optimization problems. Our implementation brings the best results when the number of cores is relatively high and the computational complexity of the fitness function is also high. The experiments with benchmark functions show that it is able to overcome its origin on the Sphere function, reaches certain thresholds faster on the Rosenbrock and Ellipsoid function, and surprisingly performs much better than the original version on the Rastrigin function.",

keywords = "CMA-ES, asynchronous algorithms, parallel algorithms",

author = "Viktor Arkhipov and Maxim Buzdalov and Shalyto, {A. A.}",

year = "2016",

month = mar,

day = "3",

doi = "10.1109/ICMLA.2015.97",

language = "English",

pages = "707--712",

booktitle = "2015 IEEE 14th International Conference on Machine Learning and Applications (ICMLA)",

publisher = "IEEE Press",

address = "United States of America",

note = "IEEE 14th International Conference on Machine Learning and Applications (ICMLA) ; Conference date: 09-12-2015 Through 11-12-2015",

}

Arkhipov, V, Buzdalov, M & Shalyto, AA 2016, An Asynchronous Implementation of the Limited Memory CMA-ES. in 2015 IEEE 14th International Conference on Machine Learning and Applications (ICMLA). IEEE Press, pp. 707-712, IEEE 14th International Conference on Machine Learning and Applications (ICMLA), Miami, Florida, United States of America, 09 Dec 2015. https://doi.org/10.1109/ICMLA.2015.97

TY - GEN

T1 - An Asynchronous Implementation of the Limited Memory CMA-ES

AU - Arkhipov, Viktor

AU - Buzdalov, Maxim

AU - Shalyto, A. A.

PY - 2016/3/3

Y1 - 2016/3/3

N2 - We present our asynchronous implementation of the LM-CMA-ES algorithm, which is a modern evolution strategy for solving complex large-scale continuous optimization problems. Our implementation brings the best results when the number of cores is relatively high and the computational complexity of the fitness function is also high. The experiments with benchmark functions show that it is able to overcome its origin on the Sphere function, reaches certain thresholds faster on the Rosenbrock and Ellipsoid function, and surprisingly performs much better than the original version on the Rastrigin function.

AB - We present our asynchronous implementation of the LM-CMA-ES algorithm, which is a modern evolution strategy for solving complex large-scale continuous optimization problems. Our implementation brings the best results when the number of cores is relatively high and the computational complexity of the fitness function is also high. The experiments with benchmark functions show that it is able to overcome its origin on the Sphere function, reaches certain thresholds faster on the Rosenbrock and Ellipsoid function, and surprisingly performs much better than the original version on the Rastrigin function.

KW - CMA-ES

KW - asynchronous algorithms

KW - parallel algorithms

UR - https://www.scopus.com/record/display.uri?eid=2-s2.0-84969718862&origin=resultslist

U2 - 10.1109/ICMLA.2015.97

DO - 10.1109/ICMLA.2015.97

M3 - Conference Proceeding (Non-Journal item)

SP - 707

EP - 712

BT - 2015 IEEE 14th International Conference on Machine Learning and Applications (ICMLA)

PB - IEEE Press

T2 - IEEE 14th International Conference on Machine Learning and Applications (ICMLA)

Y2 - 9 December 2015 through 11 December 2015

ER -

An Asynchronous Implementation of the Limited Memory CMA-ES

Abstract

Conference

Keywords

Access to Document

Permanent link

Other files and links

Fingerprint

Cite this