Tree-structured decomposition and adaptation in MOEA/D

Hanwei Zhang; Aimin Zhou

doi:10.1007/978-3-319-99253-2_29

Tree-structured decomposition and adaptation in MOEA/D

Hanwei Zhang
, Aimin Zhou^*

^*Corresponding author for this work

School of Computer Science and Technology

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

3 Scopus citations

Abstract

The multiobjective evolutionary algorithm based on decomposition (MOEA/D) converts a multiobjective optimization problem (MOP) into a set of simple subproblems, and deals with them simultaneously to approximate the Pareto optimal set (PS) of the original MOP. Normally in MOEA/D, a set of weight vectors are predefined and kept unchanged during the search process. In the last few years, it has been demonstrated in some cases that a set of predefined subproblems may fail to achieve a good approximation to the Pareto optimal set. The major reason is that it is usually unable to define a proper set of subproblems, which take full consideration of the characteristics of the MOP beforehand. Therefore, it is imperative to develop a way to adaptively redefine the subproblems during the search process. This paper proposes a tree-structured decomposition and adaptation (TDA) strategy to achieve this goal. The basic idea is to use a tree structure to decompose the search domain into a set of subdomains that are related with some subproblems, and adaptively maintain these subdomains by analyzing the search behaviors of MOEA/D in these subdomains. The TDA strategy has been applied to a variety of test instances. Experimental results show the advantages of TDA on improving MOEA/D in dealing with MOPs with different characteristics.

Original language	English
Title of host publication	Parallel Problem Solving from Nature – PPSN XV - 15th International Conference, 2018, Proceedings
Editors	Carlos M. Fonseca, Nuno Lourenco, Penousal Machado, Luis Paquete, Darrell Whitley, Anne Auger
Publisher	Springer Verlag
Pages	359-371
Number of pages	13
ISBN (Print)	9783319992525
DOIs	https://doi.org/10.1007/978-3-319-99253-2_29
State	Published - 2018
Event	15th International Conference on Parallel Problem Solving from Nature, PPSN 2018 - Coimbra, Portugal Duration: 8 Sep 2018 → 12 Sep 2018

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11101 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	15th International Conference on Parallel Problem Solving from Nature, PPSN 2018
Country/Territory	Portugal
City	Coimbra
Period	8/09/18 → 12/09/18

Access to Document

10.1007/978-3-319-99253-2_29

Cite this

Zhang, H., & Zhou, A. (2018). Tree-structured decomposition and adaptation in MOEA/D. In C. M. Fonseca, N. Lourenco, P. Machado, L. Paquete, D. Whitley, & A. Auger (Eds.), Parallel Problem Solving from Nature – PPSN XV - 15th International Conference, 2018, Proceedings (pp. 359-371). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11101 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-99253-2_29

Zhang, Hanwei ; Zhou, Aimin. / Tree-structured decomposition and adaptation in MOEA/D. Parallel Problem Solving from Nature – PPSN XV - 15th International Conference, 2018, Proceedings. editor / Carlos M. Fonseca ; Nuno Lourenco ; Penousal Machado ; Luis Paquete ; Darrell Whitley ; Anne Auger. Springer Verlag, 2018. pp. 359-371 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{6348550ba85f4f69bcf1260dfd67a99f,

title = "Tree-structured decomposition and adaptation in MOEA/D",

abstract = "The multiobjective evolutionary algorithm based on decomposition (MOEA/D) converts a multiobjective optimization problem (MOP) into a set of simple subproblems, and deals with them simultaneously to approximate the Pareto optimal set (PS) of the original MOP. Normally in MOEA/D, a set of weight vectors are predefined and kept unchanged during the search process. In the last few years, it has been demonstrated in some cases that a set of predefined subproblems may fail to achieve a good approximation to the Pareto optimal set. The major reason is that it is usually unable to define a proper set of subproblems, which take full consideration of the characteristics of the MOP beforehand. Therefore, it is imperative to develop a way to adaptively redefine the subproblems during the search process. This paper proposes a tree-structured decomposition and adaptation (TDA) strategy to achieve this goal. The basic idea is to use a tree structure to decompose the search domain into a set of subdomains that are related with some subproblems, and adaptively maintain these subdomains by analyzing the search behaviors of MOEA/D in these subdomains. The TDA strategy has been applied to a variety of test instances. Experimental results show the advantages of TDA on improving MOEA/D in dealing with MOPs with different characteristics.",

author = "Hanwei Zhang and Aimin Zhou",

note = "Publisher Copyright: {\textcopyright} Springer Nature Switzerland AG 2018.; 15th International Conference on Parallel Problem Solving from Nature, PPSN 2018 ; Conference date: 08-09-2018 Through 12-09-2018",

year = "2018",

doi = "10.1007/978-3-319-99253-2\_29",

language = "英语",

isbn = "9783319992525",

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

publisher = "Springer Verlag",

pages = "359--371",

editor = "Fonseca, \{Carlos M.\} and Nuno Lourenco and Penousal Machado and Luis Paquete and Darrell Whitley and Anne Auger",

booktitle = "Parallel Problem Solving from Nature – PPSN XV - 15th International Conference, 2018, Proceedings",

address = "德国",

}

Zhang, H & Zhou, A 2018, Tree-structured decomposition and adaptation in MOEA/D. in CM Fonseca, N Lourenco, P Machado, L Paquete, D Whitley & A Auger (eds), Parallel Problem Solving from Nature – PPSN XV - 15th International Conference, 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11101 LNCS, Springer Verlag, pp. 359-371, 15th International Conference on Parallel Problem Solving from Nature, PPSN 2018, Coimbra, Portugal, 8/09/18. https://doi.org/10.1007/978-3-319-99253-2_29

Tree-structured decomposition and adaptation in MOEA/D. / Zhang, Hanwei; Zhou, Aimin.
Parallel Problem Solving from Nature – PPSN XV - 15th International Conference, 2018, Proceedings. ed. / Carlos M. Fonseca; Nuno Lourenco; Penousal Machado; Luis Paquete; Darrell Whitley; Anne Auger. Springer Verlag, 2018. p. 359-371 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11101 LNCS).

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

TY - GEN

T1 - Tree-structured decomposition and adaptation in MOEA/D

AU - Zhang, Hanwei

AU - Zhou, Aimin

N1 - Publisher Copyright: © Springer Nature Switzerland AG 2018.

PY - 2018

Y1 - 2018

N2 - The multiobjective evolutionary algorithm based on decomposition (MOEA/D) converts a multiobjective optimization problem (MOP) into a set of simple subproblems, and deals with them simultaneously to approximate the Pareto optimal set (PS) of the original MOP. Normally in MOEA/D, a set of weight vectors are predefined and kept unchanged during the search process. In the last few years, it has been demonstrated in some cases that a set of predefined subproblems may fail to achieve a good approximation to the Pareto optimal set. The major reason is that it is usually unable to define a proper set of subproblems, which take full consideration of the characteristics of the MOP beforehand. Therefore, it is imperative to develop a way to adaptively redefine the subproblems during the search process. This paper proposes a tree-structured decomposition and adaptation (TDA) strategy to achieve this goal. The basic idea is to use a tree structure to decompose the search domain into a set of subdomains that are related with some subproblems, and adaptively maintain these subdomains by analyzing the search behaviors of MOEA/D in these subdomains. The TDA strategy has been applied to a variety of test instances. Experimental results show the advantages of TDA on improving MOEA/D in dealing with MOPs with different characteristics.

AB - The multiobjective evolutionary algorithm based on decomposition (MOEA/D) converts a multiobjective optimization problem (MOP) into a set of simple subproblems, and deals with them simultaneously to approximate the Pareto optimal set (PS) of the original MOP. Normally in MOEA/D, a set of weight vectors are predefined and kept unchanged during the search process. In the last few years, it has been demonstrated in some cases that a set of predefined subproblems may fail to achieve a good approximation to the Pareto optimal set. The major reason is that it is usually unable to define a proper set of subproblems, which take full consideration of the characteristics of the MOP beforehand. Therefore, it is imperative to develop a way to adaptively redefine the subproblems during the search process. This paper proposes a tree-structured decomposition and adaptation (TDA) strategy to achieve this goal. The basic idea is to use a tree structure to decompose the search domain into a set of subdomains that are related with some subproblems, and adaptively maintain these subdomains by analyzing the search behaviors of MOEA/D in these subdomains. The TDA strategy has been applied to a variety of test instances. Experimental results show the advantages of TDA on improving MOEA/D in dealing with MOPs with different characteristics.

UR - https://www.scopus.com/pages/publications/85053630215

U2 - 10.1007/978-3-319-99253-2_29

DO - 10.1007/978-3-319-99253-2_29

M3 - 会议稿件

AN - SCOPUS:85053630215

SN - 9783319992525

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 359

EP - 371

BT - Parallel Problem Solving from Nature – PPSN XV - 15th International Conference, 2018, Proceedings

A2 - Fonseca, Carlos M.

A2 - Lourenco, Nuno

A2 - Machado, Penousal

A2 - Paquete, Luis

A2 - Whitley, Darrell

A2 - Auger, Anne

PB - Springer Verlag

T2 - 15th International Conference on Parallel Problem Solving from Nature, PPSN 2018

Y2 - 8 September 2018 through 12 September 2018

ER -

Zhang H, Zhou A. Tree-structured decomposition and adaptation in MOEA/D. In Fonseca CM, Lourenco N, Machado P, Paquete L, Whitley D, Auger A, editors, Parallel Problem Solving from Nature – PPSN XV - 15th International Conference, 2018, Proceedings. Springer Verlag. 2018. p. 359-371. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-99253-2_29

Tree-structured decomposition and adaptation in MOEA/D

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this