Paper Info
Title
Fitness Landscape Analysis In The Optimization Of Coefficients Of Curve Parametrizations
Abstract
Parametric representations of geometric objects, such as curves or surfaces, may have unnecessarily huge integer coefficients. Our goal is to search for an alternative parametric representation of the same object with significantly smaller integer coefficients. We have developed and implemented an evolutionary algorithm that is able to find solutions to this problem in an efficient as well as robust way.In this paper we analyze the fitness landscapes associated with this evolutionary algorithm. We here discuss the use of three different strategies that are used to evaluate and order partial solutions. These orderings lead to different landscapes of combinations of partial solutions in which the optimal solutions are searched. We see that the choice of this ordering strategy has a huge influence on the characteristics of the resulting landscapes, which are in this paper analyzed using a set of metrics, and also on the quality of the solutions that can be found by the subsequent evolutionary search.
Year
DOI
Venue
2017
10.1007/978-3-319-74718-7_56
COMPUTER AIDED SYSTEMS THEORY - EUROCAST 2017, PT I
Field
DocType
Volume
Integer,Mathematical optimization,Fitness landscape,Evolutionary algorithm,Computer science,Theoretical computer science,Parametric statistics
Conference
10671
ISSN
Citations 
PageRank 
0302-9743
0
0.34
References 
Authors
3
2
Name
Order
Citations
PageRank
Stephan M. Winkler114022.90
J. Rafael Sendra262168.33