1545-5963 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCBB.2017.2701367, IEEE/ACM
Transactions on Computational Biology and Bioinformatics
1
A Grouping Particle Swarm Optimizer with
Personal-Best-Position Guidance for Large Scale
Optimization
Weian Guo, Member, IEEE, Chengyong Si, Member, IEEE, Yu Xue, Member, IEEE, Yanfen Mao,
Member, IEEE, Lei Wang, Member, IEEE, and Qidi Wu, Senior Member, IEEE
Abstract—Particle Swarm Optimization (PSO) is a popular
algorithm which is widely investigated and well implemented in
many areas. However, the canonical PSO does not perform well
in population diversity maintenance so that usually leads to a
premature convergence or local optima. To address this issue, we
propose a variant of PSO named Grouping PSO with Personal-
Best-Position (P
best
) Guidance (GPSO-PG) which maintains the
population diversity by preserving the diversity of exemplars.
On one hand, we adopt uniform random allocation strategy
to assign particles into different groups and in each group the
losers will learn from the winner. On the other hand, we employ
personal historical best position of each particle in social learning
rather than the current global best particle. In this way, the
exemplars diversity increases and the effect from the global
best particle is eliminated. We test the proposed algorithm to
the benchmarks in CEC 2008 and CEC 2010, which concern
the large scale optimization problems (LSOPs). By comparing
several current peer algorithms, GPSO-PG exhibits a competitive
performance to maintain population diversity and obtains a
satisfactory performance to the problems.
Index Terms—Particle Swarm Optimization; Population Diver-
sity Maintenance; Premature Convergence; Random Allocation;
Personal-Best-Position Guidance; Large Scale Optimization
I. INTRODUCTION
In current decades, particle swarm optimization (PSO), as a
popular Swarm Intelligence (SI), draws worldwide attentions
due to its simplicity and efficiency in optimization [1], [2],
[3]. Both studies and applications of PSO demonstrate that
the algorithm is competitive to handle optimization problems
and plays a very active role to deal with practical cases such as
image processing [4], flow-shop scheduling [5], array synthesis
[6], power system [7], mechanical design [8] and many others
[9]. In the mechanism of canonical PSO, the personal historical
best position named p
best
and the global best position named
g
best
play the role to provide all particles information for
convergence. Nevertheless, in canonical PSO, if the g
best
gets
trapped into a local optimal position, it is difficult for the whole
swarm to get rid of it. Especially, for the problems which
demand population diversity, such as Large Scale Optimization
Problems (LSOPs), PSO’s performance deteriorates much.
Weian Guo and Yanfen Mao are with Sino-German College of Applied
Sciences, Tongji University, Shanghai, 201804, China.
Chenyong Si (Corresponding Author) is with Shanghai-Hamburg College,
University of Shanghai for Science and Technology, Shanghai, 200093, China
Yu Xue is with School of Computer and Software, Nanjing University of
Information Science and Technology, Nanjing, Jiangsu, 210044, China
Lei Wang and Qidi Wu are with School of Electronics and Information
Engineering, Tongji University, Shanghai, 201804, China.
The reasons are mainly summed up in two aspects. One is
that a high dimension increases problem’s complexity which
increases difficulty for algorithms to break away the local
optima. The other one is that, for the searching space, both
size and complexity increase, which leads to a difficulty that
algorithms should efficiently explore the searching space by
only limited computational resources.
However, on the other hand, an unsuitable population diver-
sity may cause a poor ability in convergence. Too much weight
on diversity maintenance will slow down the convergence
speed or cause a convergence stagnation. To address the issues,
a proper method for population diversity maintenance is much
advisable. To date, many scholars investigate various ways for
PSO. The popular ways of preserving population diversity are
summarized as follows:
1) The first one is to design adaptive parameters for canon-
ical PSO. In PSO, inertia weight ω is used to balance
effects of convergence and diversity maintenance. A
large value of ω is helpful for global exploration, while
a small value of ω works on the exploitation. Hence
it is a natural idea to design a varying ω to enhance
the adaptive ability for PSO to balance convergence
and diversity maintenance. Several kinds of linearly
decreasing inertia weight methods are proposed in [1].
Fuzzy strategy is also employed in [2] to adjust the
inertia weight. In addition to the tuning on inertia
weight, by analyzing the convergence behavior of PSO,
a PSO variant with a constriction factor is introduced by
Clerc and Kennedy [10], where the constriction factor
aids to control over algorithm’s convergence ability and
diversity maintenance. However, the performance of
PSO is sensitive to different parameters and therefore
the method by parameter tuning is a time-consuming
task for different problems.
2) The second one is to carry out the researches on the
operators of PSO. In [11], a variable named current
critical value is used to depict the criticality of a swarm’s
status for PSO. If the critical values for some particles
are larger than a preset criticality limit, a relocating
operator will be activated to disperse the particles in
order to prevent particles from being too close. Therefore
the swarm diversity can be maintained. In [12], an
idea inspired by electrostatics is proposed, which sets
particles as charged ones and uncharged ones. For two
剩余12页未读，继续阅读
资源评论
