Forking Genetic Algorithm with Blocking and Shrinking Modes (fGA)

grandgoatAI and Robotics

Oct 23, 2013 (3 years and 9 months ago)

132 views

Forking Genetic Algorithm with Blocking
and Shrinking Modes (fGA)
Shigeyoshi Tsutsui Yoshiji Fujimoto
Department of Management and Printing and Reprographic Systems
Information Science Product Development Laboratory
HANNAN University SHARP Corporation
5-4-33 Amamihigashi, Matsubara, 492 Minosho, Yamatokoriyama,
Osaka 580 Japan Nara 639-11 Japan
1 Introduction
There are many GA-hard problems that are
difficult to solve by the traditional GAs such as
problems with multi-modal and deceptive evalu-
ation functions [4] [15]. Many kinds of modified
GAs that are aimed to solve these problems are
pr oposed such as CHC[ 2], mGA[ 5],
GENITOR[13] and Niche Method[4].
In this paper, we propose a new type of GA,
that is, the forking Genetic Algorithm (fGA). The
fGA is designed to solve such problems as have
multi-modal evaluation functions with many lo-
cal optimal points. This GA evolves multi-popu-
lations. In conventional GAs with multi-popula-
tions [1] [6] [10], each population is indepen-
dently evolved in the same genetic operations.
They maintain and enrich diversity gained by ge-
netic drift through immigration of individuals be-
tween populations [6] [10]. The distinguishing
feature of the fGA is that it has one parent popu-
lation with a blocking mode and one or more
child populations with a shrinking mode as a re-
sult of population forking. Each population takes
a different role in optimizing tasks. That is, each
population is responsible for searching for non-
overlapping sub-areas in the search space.
Genetic operators of the fGA, the process of
the population forking and empirical results are
described in the following sections.
2 Generational, Overlapping and Best
N Selection Methods
In modified GAs, there are generational evo-
lution where genetic operations are applied to
whole individuals in the population simulta-
neously and steady state evolution where genetic
operations are applied to individuals, one by one
[12]. Usually, in generational evolution, there is
no overlapping of parent and offspring individu-
als and in steady state evolution, there is overlap-
Abstract
In this paper, we propose a new type of
multi-population GA, that is, the forking
Genetic Algorithm (fGA). The fGA is de-
signed to solve multi-modal problems which
are difficult to solve by the traditional GAs
because of the many local optimums. The
fGA has the following features:
(1) generational and overlapping evolution
strategy,
(2) selective crossover and high mutation
with the best N selection, and
(3) multi-population with one parent popula-
tion with blocking mode and one or
more child populations with shrinking
mode.
We take two problems as test functions. One
is a FM SoundÕs parameter identification
problem and the other is OliverÕs 30 City
Travel Salesperson Problem. The results of
experiments with a fixed number of trials
that include a number of times to find an op-
timal solution and an average value of evalu-
ation function, show that the fGA outper-
forms the standard GA.
type diversity.
The future studies of the fGA are as follows.
(1) Tuning of control parameters related to the
salient schema and applications of the other
genetic operators such as uniform cross-
over.
(2) Corroboration of effectiveness of the fGA
on a wide variety of multi-modal problems
(3) Implementation methods of the fGA on par-
allel processors.
References
[1] J.P. Cohoon, W.N. Martin, and D.S.
Richards, "A Multi-Population Genetic Al-
gorithm for Solving the K-Partition Problem
on Hyper-Cubes" Proc. of the Fourth ICGA,
pp. 244-248, July, 1991.
[2] L. J. Eshelman, "5IF$)$"EBQUJWF4FBSDI
"MHPSJUIN)PXUP)BWF4BGF4FBSDI8IFO
&OHBHJOHJO/POUSBEJUJPOBM(FOFUJD3FDPN
CJOBUJPO, Foundations of Genetic Algo-
rithms edited by Gregory J.E. Rawlins, Mor-
gan Koufmann, pp. 265-283, 1991.
[3] L.J. Eshelman and J.D. Shaffer, "1SFWFOUJOH
1SFNBUVSF $POWFSHFODF JO (FOFUJD"MHP
SJUIN CZ 1SFWFOUJOH *ODFTU, Proc. of the
Fourth ICGA, pp. 115-122, July, 1991.
[4] D.E. Goldberg, Genetic Algorithms in
Search, Optimization and Machine Learning,
Addison-Wesley Publishing Company, Inc.,
1989.
[5] D.E. Goldberg, B. Korb and K. Deb, ".FTTZ
(FOFUJD"MHPSJUINT.PUJWBUJPO "OBMZTJT
BOE'JSTU3FTVMUT, Complex Systems, vol. 3,
pp. 493-530, 1989.
[6] Martina Gorges-Schleuter, "ASPARAGOS
An Asynchronous Parallel Genetic Optimiza-
tion Strategy", Proc. of the Third ICGA, pp.
422-427, June 1989.
[7] J.J. Grefenstette, L. Davis and D. Cerys,
"GENESIS and OOGA: Two GA Systems"
TSP Publication, Melrose, MA, USA, 1991.
[8] I.M. Oliver, D.J. Smith and J.R.C. Holland,
""4UVEZPG1FSNVUBUJPO$SPTTPWFS0QFSB
UPSTPOUIF5SBWFMJOH4BMFTNBO1SPCMFN,
Proc. of the Second ICGA, pp. 224-230, July
1987.
[9] J. D. Schaffer, R. A. Caruana, L. J.
Eshelman, and R. Das, ""4UVEZPG$POUSPM
1BSBNFUFST"GGFDUJOH0OMJOF1FSGPSNBODFPG
(FOFUJD"MHPSJUINTGPS'VODUJPO0QUJNJ[FS,
Proc. of the Third ICGA, pp. 52-60, June
1989.
[10] T. Starkweather, D. Whitley and K Mathias,
"0QUJNJ[BUJPO6TJOH%JTUSJCVUFE(FOFUJD"M
HPSJUINT, Parallel Problem Solving from
Nature, Proc. of First workshop, PPSN 1, pp.
109-116, Springer-Verlag, Dortmund, FRG,
October 1990.
[11] T. Starkweather, S. McDaniel, K Mathias,
D. Whitley and C. Whitley, "A Comparison
of Genetic Sequencing Operators", Proc. of
the Fourth ICGA, pp. 69-76, July, 1991.
[12] G. Syswerda, "A Study of Reproduction in
Generational and Steady-State Genetic Algo-
rithms", Foundations of Genetic Algorithms
edited by Gregory J.E. Rawlins, Morgan
Koufmann, pp. 94-101, 1991.
[13] D. Whitley, "The GENITOR Algorithm and
Selection Pressure: Why Rank-Based Alloca-
tion of Reproductive Trials is Best", Proc. of
the Third ICGA, pp. 116-121, June 1989.
[14] D. Whitley, T. Starkweather and D'Ann
Fuquay, "Scheduling Problems and Travel-
ing Salesmen: The Genetic Edge Recombina-
tion Operator", Proc. of the Third ICGA, pp.
133-140, June 1989.
[15] D. Whitley, "Fundamental Principles of De-
ception in Genetic Search", Foundations of
Genetic Algorithms edited by Gregory J.E.
Rawlins, Morgan Koufmann, pp. 221-241,
1991.