|
Digital
Library of the European Council for Modelling
and Simulation |
Title: |
Backbone Strategy For Unconstrained Continuous
Optimization |
Authors: |
Michael
Feldmeier, Thomas Husslein |
Published in: |
(2017).ECMS 2017 Proceedings
Edited by: Zita Zoltay Paprika, Péter Horák, Kata Váradi, Péter Tamás
Zwierczyk, Ágnes Vidovics-Dancs, János Péter Rádics European Council for Modeling and Simulation. doi:10.7148/2017 ISBN:
978-0-9932440-4-9/ ISBN:
978-0-9932440-5-6 (CD) 31st European Conference on Modelling and
Simulation, Budapest, Hungary, May 23rd
– May 26th, 2017 |
Citation
format: |
Michael
Feldmeier, Thomas Husslein (2017). Backbone Strategy For Unconstrained
Continuous Optimization, ECMS 2017 Proceedings Edited by: Zita Zoltay
Paprika, Péter Horák, Kata Váradi, Péter Tamás Zwierczyk, Ágnes
Vidovics-Dancs, János Péter Rádics European Council for Modeling and
Simulation. doi: 10.7148/2017-0529 |
DOI: |
https://doi.org/10.7148/2017-0529 |
Abstract: |
Backbones
in optimization problems are structures within the decision variables that
are common to all global optima. Identifying those backbones in a
deterministic manner is at least as hard as solving the original problem to
optimality because all optimal solutions to the problem have to be known. A
number of different algorithms have been proposed which use heuristically
determined backbones to speed up discrete combinatorial optimization
algorithms by eliminating these backbones and thus reducing the
dimensionality of the optimization problem to be solved in each step. In this
paper we extend the concept of backbones to real-valued optimization. We
propose a definition of such backbones and introduce means to identify them
and determine their value by the use of a genetic algorithm. We compare the
performance of the resulting algorithm with an ordinary optimization
procedure on a widely used nonlinear and unconstrained optimization
benchmark. We observe that our backbone strategy is superior in terms of both
convergence speed and quality of the resulting solutions. Limitations of this
first approach and ideas how to resolve them in future work are considered. |
Full
text: |