TY - GEN
T1 - Fitness landscape analysis in the optimization of coefficients of curve parametrizations
AU - Winkler, Stephan M.
AU - Sendra, J. Rafael
N1 - Publisher Copyright:
© Springer International Publishing AG 2018.
Copyright:
Copyright 2018 Elsevier B.V., All rights reserved.
PY - 2018
Y1 - 2018
N2 - 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.
AB - 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.
UR - http://www.scopus.com/inward/record.url?scp=85041827354&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-74718-7_56
DO - 10.1007/978-3-319-74718-7_56
M3 - Conference contribution
AN - SCOPUS:85041827354
SN - 9783319747170
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 464
EP - 472
BT - Computer Aided Systems Theory – EUROCAST 2017 - 16th International Conference, Revised Selected Papers
A2 - Moreno-Diaz, Roberto
A2 - Quesada-Arencibia, Alexis
A2 - Pichler, Franz
PB - Springer
T2 - 16th International Conference on Computer Aided Systems Theory, EUROCAST 2017
Y2 - 19 February 2017 through 24 February 2017
ER -