TY - GEN
T1 - A fast objective reduction algorithm based on dominance structure for many objective optimization
AU - Gu, Fangqing
AU - Liu, Hai Lin
AU - CHEUNG, Yiu Ming
N1 - Funding Information:
This work was supported by the National Natural Science Foundation of China under Grants 61672444, 61673121 and 61703108, in part by the Natural Science Foundation of Guangdong Province under Grant 2017A030310467, and in part by the Projects of Science and Technology of Guangzhou under Grant 201508010008, the SZSTI Grant: JCYJ20160531194006833, the Faculty Research Grant of Hong Kong Baptist University (HKBU) under Project: FRG2/16-17/051, and the MPCF Project of Knowledge Transfer Office in HKBU: MPCF-004-2017/18.
PY - 2017
Y1 - 2017
N2 - The performance of the most existing classical evolutionary multiobjective optimization (EMO) algorithms, especially for Pareto-based EMO algorithms, generally deteriorates over the number of objectives in solving many-objective optimization problems (MaOPs), in which the number of objectives is greater than three. Objective reduction methods that transform an MaOP into the one with few objectives, are a promising way for solving MaOPs. The dominance-based objective reduction methods, e.g. k-EMOSS and δ-MOSS, omitting an objective while preserving the dominant structure of the individuals as much as possible, can achieve good performance. However, these algorithms have higher computational complexity. Therefore, this paper presents a novel measure for measuring the capacity of preserving the dominance structure of an objective set, i.e., the redundancy of an objective to an objective set. Subsequently, we propose a fast algorithm to find a minimum set of objectives preserving the dominance structure as much as possible. We compare the proposed algorithm with its counterparts on eleven test instances. Numerical studies show the effectiveness of the proposed algorithm.
AB - The performance of the most existing classical evolutionary multiobjective optimization (EMO) algorithms, especially for Pareto-based EMO algorithms, generally deteriorates over the number of objectives in solving many-objective optimization problems (MaOPs), in which the number of objectives is greater than three. Objective reduction methods that transform an MaOP into the one with few objectives, are a promising way for solving MaOPs. The dominance-based objective reduction methods, e.g. k-EMOSS and δ-MOSS, omitting an objective while preserving the dominant structure of the individuals as much as possible, can achieve good performance. However, these algorithms have higher computational complexity. Therefore, this paper presents a novel measure for measuring the capacity of preserving the dominance structure of an objective set, i.e., the redundancy of an objective to an objective set. Subsequently, we propose a fast algorithm to find a minimum set of objectives preserving the dominance structure as much as possible. We compare the proposed algorithm with its counterparts on eleven test instances. Numerical studies show the effectiveness of the proposed algorithm.
KW - Evolutionary algorithm
KW - Many-objective optimization
KW - Objective reduction
UR - http://www.scopus.com/inward/record.url?scp=85034269878&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-68759-9_22
DO - 10.1007/978-3-319-68759-9_22
M3 - Conference proceeding
AN - SCOPUS:85034269878
SN - 9783319687582
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 260
EP - 271
BT - Simulated Evolution and Learning - 11th International Conference, SEAL 2017, Proceedings
A2 - Li, Xiaodong
A2 - Zhang, Mengjie
A2 - Zhang, Qingfu
A2 - Middendorf, Martin
A2 - Tan, Kay Chen
A2 - Tan, Ying
A2 - Jin, Yaochu
A2 - Shi, Yuhui
A2 - Tang, Ke
PB - Springer Verlag
T2 - 11th International Conference on Simulated Evolution and Learning, SEAL 2017
Y2 - 10 November 2017 through 13 November 2017
ER -