Knowledge-Based Systems 125 (2017) 108–115
Contents lists available at ScienceDirect
Knowle dge-Base d Systems
journal homepage: www.elsevier.com/locate/knosys
An improvement decomposition-based multi-objective evolutionary
algorithm with uniform design
Cai Dai
∗
, Xiujuan Lei
College of Computer Science, Shaanxi Normal University, Xi’an, 710119, China
a r t i c l e i n f o
Article history:
Received 13 July 2016
Revised 8 January 2017
Accepted 25 March 2017
Available online 30 March 2017
Keywords:
Multi-objective optimization
Decomposition
Uniform design
Descent direction
a b s t r a c t
How to quickly find a set of solutions with good diversity and convergence is the main goal of multi-
objective optimization evolutionary algorithms (MOEAs). In this paper, a crossover operator based on
uniform design and selection strategy based on decomposition is designed to help MOEAs to improve the
search efficiency, and an improvement decomposition-based multi-objective evolutionary algorithm with
uniform design is proposed. Firstly, a multi-objective problem is transformed into a set of single problems
based on a set of direction vectors, and all single problems are optimized simultaneously. Secondly, a
crossover operator based on uniform design which can search decision space along the descent (ascent)
directions is designed to improve the search efficiency of the algorithm. Thirdly, in order to improve
the convergence performance of the algorithm, a sub-population strategy is used to optimize each sub-
problem. Moreover, a selection strategy is designed to help the crossover operators to balance between
the global searching and the local searching. Comparing with some efficient state-of-the-art algorithms,
e.g., NSGAII and MOEA/D, on some benchmark functions, the proposed algorithm is able to find a set of
solutions with better diversity and convergence.
©2017 Elsevier B.V. All rights reserved.
1.
Introduction
In real-world applications, there are many problems need to
simultaneously optimize multiple objectives which are typically
characterized by conflicting objectives. These problems are called
as multi-objective optimization problems (MOPs). The goal of solv-
ing MOPs is to find a set of optimal solutions which are called as
Pareto optimal solutions (PS) not a single optimal solution. The im-
age of all Pareto optimal solutions is known as the Pareto front
(PF). Because the number of PF may be infinite, it is impractical to
obtain all the Pareto optimal solutions. Thus, the important goal of
solving MOPs is to find out a set of solutions with good diversity
and convergence.
Currently, multi-objective evolutionary algorithms (MOEAs)
which use the strategy of the population evolution to obtain a
set of solutions in a run are an effective method to solve MOPs.
MOEAs can well deal with some complex problems which are
characterized with discontinuity, multimodality and nonlinearity
[1] . Nowadays, many MOEAs [2–24,39,41–45] with good perfor-
mance have been proposed, such as multi-objective genetic algo-
rithms [3] , multi-objective particle swarm optimization algorithms
∗
Corresponding author.
E-mail addresses: cdai0320@snnu.edu.cn , daicai8403@hotmail.com (C. Dai),
xjlei@snnu.edu.cn (X. Lei).
[5,6,7,8] , multi-objective differential evolution algorithms [8,9] ,
multi-objective immune clone algorithms [10] , group search opti-
mizer [11] , evolutionary algorithms based on decomposition [12–
15,44–45]
and hybrid algorithms [6,24] . Moreover, many MOEAs
are used to solve many applications [25–28] .
Here, we are interested in how to improve the performance
of MOEAs. According to the literature [29] , we think that the
crossover operators directly affect the search efficiency and good
selection strategies can help crossover operators to balance be-
tween exploration and exploitation. Thus, the main goal of this
paper is to study that uses crossover operators with appropriate
selection strategy to improve the performance of MOEAs. Accord-
ing to this idea, an improved decomposition-based multi-objective
evolutionary algorithm is designed. Firstly, a decomposition ap-
proach is used to transform MOPs into a set of sub-problems,
and these sub-problems are simultaneously optimized. Secondly, a
crossover operator based on uniform design which can search deci-
sion space along the descent directions is proposed to enhance the
search efficiency. Thirdly, a selection strategy based on decomposi-
tion and sub-population strategy is proposed to help the crossover
operators to balance between exploration and exploitation. More-
over, an updated strategy based on decomposition is designed to
maintain the diversity of obtained solutions and help solutions to
converge to the true Pareto optimal solutions.
http://dx.doi.org/10.1016/j.knosys.2017.03.021
0950-7051/© 2017 Elsevier B.V. All rights reserved.