TY - GEN
T1 - Combining two-phase local search with multi-objective ant colony optimization
AU - Leung, Chun Wa
AU - Ng, Sin Chun
AU - Lui, Andrew K.
N1 - Publisher Copyright:
© 2018, Springer Nature Switzerland AG.
PY - 2018
Y1 - 2018
N2 - Multi-objective Ant Colony Optimization (MOACO) is a popular algorithm in solving the multi-objective combinational optimization problem. Many variants were introduced to solve different types of multi-objective optimization problem. However, MOACO does not guarantee to generate a good approximation of solution in a predefined termination time. In this paper, Two-Phase Local Search (TPLS) was introduced to cooperate with the MOACO, a two-phase strategy that creates a very good approximation of Pareto Front at the beginning of the algorithm and then further explores the Pareto Front iteratively. We also propose Iterated Local Search – Variable Neighborhood Search (ILS-VNS) as the first phase in TPLS, an iterative improvement process that allows finding improving solutions from adaptively sized neighborhood space. A series of experiments were performed to investigate the performance improvement on the solutions. At the same time, we studied the effect of Weighted Local Search (WLS) and Pareto Local Search (PLS) in the proposed algorithm. The results showed that the newly proposed algorithm obtains a larger area on hypervolume space and exhibits a significantly larger accuracy rate in obtaining the true Pareto-optimal solutions. Additionally, a statistical testing was also performed to verify the significance of the result.
AB - Multi-objective Ant Colony Optimization (MOACO) is a popular algorithm in solving the multi-objective combinational optimization problem. Many variants were introduced to solve different types of multi-objective optimization problem. However, MOACO does not guarantee to generate a good approximation of solution in a predefined termination time. In this paper, Two-Phase Local Search (TPLS) was introduced to cooperate with the MOACO, a two-phase strategy that creates a very good approximation of Pareto Front at the beginning of the algorithm and then further explores the Pareto Front iteratively. We also propose Iterated Local Search – Variable Neighborhood Search (ILS-VNS) as the first phase in TPLS, an iterative improvement process that allows finding improving solutions from adaptively sized neighborhood space. A series of experiments were performed to investigate the performance improvement on the solutions. At the same time, we studied the effect of Weighted Local Search (WLS) and Pareto Local Search (PLS) in the proposed algorithm. The results showed that the newly proposed algorithm obtains a larger area on hypervolume space and exhibits a significantly larger accuracy rate in obtaining the true Pareto-optimal solutions. Additionally, a statistical testing was also performed to verify the significance of the result.
KW - Bi-objective traveling salesman problem
KW - Hybridization
KW - Iterated local search – Variable neighborhood search
KW - Multi-objective ant colony optimization
KW - Two-phase local search
UR - http://www.scopus.com/inward/record.url?scp=85059057617&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-04179-3_50
DO - 10.1007/978-3-030-04179-3_50
M3 - Conference contribution
AN - SCOPUS:85059057617
SN - 9783030041786
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 564
EP - 576
BT - Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings
A2 - Leung, Andrew Chi Sing
A2 - Ozawa, Seiichi
A2 - Cheng, Long
T2 - 25th International Conference on Neural Information Processing, ICONIP 2018
Y2 - 13 December 2018 through 16 December 2018
ER -