Adaptive Genetic Algorithms - Computational Intelligence in Time Series Forecasting - page 321

x define some firm measures related to the GA behaviour, its setting
parameters, and operators, e.g. example the diversity indices, maximum,
average, and minimum fitness values, as system inputs
x define as system outputs the values of control parameters or of their
changes
x define the database as a collection of membership functions and the
boundary values of input and output variables
x build the rule base in which the fuzzy rules describe the relations between
the input and output variables.

The statistical data generated by the genetic algorithm concern the genotypes of
individuals of a population as well as the phenotypes related to the fitness and
other properties of individual performance for the problem to be solved. Two
typical examples for the above statistics are the

x genotypic diversity measure , representing the variations of similarity
within the genetic material (like chromosomes, alleles, etc. )
x phenotypic diversity measure ,

which mainly concerns the fitness of

chromosomes.

9.6 Concluding Remarks

In this chapter, three possibilities of adaptive versions of genetic algorithms are
presented, the first of which dynamically controls the basic tuning parameters, such
as probability of crossover, mutation and reproduction etc ., based on the on-line
measurement of GA convergence. Other methods control mainly the population
size, based either on the concept of the age of the chromosome, the age structure of
the population, or by application of the average-fitness -to- best-fitness ratio,
worst-fitness -to- best-fitness ratio, besides the mutation- and crossover-rates-based
fuzzy IF-THEN rules. The efficiencies of the various methods are demonstrated on
application examples that can be found in the corresponding publications list.

References

[1] Arabas J, Michalewicz J, and Mulawka (1994) GAVaPS - a Genetic Algorithm with

varying population size. Proc. of the 1 st IEEE Conf. on Evolutionary Computation:
73-78.

[2] Arnone S, Dell'Orto M, and Tettamanzi A (1994) Toward a fuzzy government of

genetic populations. Proc. of the 6th IEEE Conf. on Tools with the Artificial
Intelligence TAI'94, IEEE Computer Press, Los Alamitos, CA.

Baker J (1985) Adaptive selection methods for genetic algorithms. In: Proc.1 st Intl.
Conf. on Genetic Algorithms (J.J. Grefenstette, ed.): 101-111. Lawrence Erlbaum
Associates, Hillsdale , NJ.

[3]

[4] BergmannA, Burgard W, and Hemker A (1994) Adjusting parameters of genetic

algorithms by fuzzy control rules. In K.-H. Becks and D. Perret-Gallix, editors, New
Computing Techniques in Physics Research III. World Scientific Press, Singapore.