Biomedical Engineering Reference
In-Depth Information
Chapter XVII
Graph Based Evolutionary
Algorithms
Steven M. Corns
Iowa State University, USA
Daniel A. Ashlock
University of Guelph, Canada
Kenneth Mark Bryden
Iowa State University, USA
AbSTRACT
This chapter presents Graph Based Evolutionary Algorithms. Graph Based Evolutionary Algorithms
are a generic enhancement and diversity management technique for evolutionary algorithms. These
geographically inspired algorithms are different from other methods of diversity control in that they
not only control the rate of diversity loss at low runtime cost but also allow for a means to classify
evolutionary computation problems. This classification system enables users to select an algorithm a
priori that finds a satisfactory solution to their optimization problem in a relatively small number of
fitness evaluations. In addition, using the information gathered by evaluating several problems on a
collection of graphs, it becomes possible to design test suites of problems which effectively compare a
new algorithm or technique to existing methods.
INTRODUCTION
evaluations required to find an acceptable solu-
tion at exceedingly low runtime cost through
the selection of the correct graph as a population
structure. The combinatorial graph provides a
geography for the evolving population of solu-
tions. Different graphs yield different levels of
diversity preservation. The level of preservation
needed for effective solution is problem specific.
Graph based evolutionary algorithms (GBEAs)
are a diversity management technique for evo-
lutionary algorithms which places the members
of the evolving population on the vertices of a
combinatorial graph. They also provide a generic
improvement by reducing the number of fitness
Search WWH ::
Custom Search