Download:
by R. Schilling, W. Haase, J. Periaux, H. Baier, G. Bugeda (eds, Chris M. Varcol, Michael T. M. Emmerich, Chris M. Varcol, Michael T. M. Emmerich
http://www.liacs.nl/~emmerich/pdf/VE05.pdf
Add To MetaCart
Abstract:
Abstract. Metamodel-Assisted Evolution Strategies (MAES) are proposed as a new approach for simulator-based optimization in the domain of Electromagnetic Compatibility (EMC) design. In the given application domain function evaluations tend to be very time consuming and thus techniques like metamodeling are to be considered in order to accelerate evolutionary optimization strategies. The MAES partially replaces the time consuming EMC computer model by so-called gaussian random field metamodels. These metamodels can be used for interpolation of objective function for new input vectors. All interpolations are based on values from a database of previously obtained function evaluations, that might stem from previous optimization runs or that are sampled during the course of optimization. Gaussian random field models not only allow for a fast prediction of the objective function value, but also augment the predicted result with a confidence value. This value can be used to improve the performance of the MAES, whenever difficult nonlinear problems need to be solved. The results indicates that the MAES is both a versatile and effective tool for automatically improving the quality of devices in electromagnetic compatibility design. 1
Citations
|
172
|
A Survey of Evolution Strategies
– Back, Hoffmeister, et al.
- 1991
|
|
144
|
em Design and Analysis of Computer Experiments
– Sacks, Welch, et al.
- 1989
|
|
79
|
Evolution strategies – A comprehensive introduction
– Beyer, Schwefel
- 2002
|
|
28
|
A comprehensive survey of fitness approximation in evolutionary computation
– Jin
- 2005
|
|
28
|
Direct search methods: Then and now
– LEWIS, TORCZON, et al.
- 2000
|
|
26
|
Metamodeling techniques for evolutionary optimization of computationally expensive problems: promises and limitations
– El-Beltagy, Nair, et al.
- 1999
|
|
24
|
Accelerating the convergence of evolutionary algorithms by fitness landscape approximation
– Ratle
- 1998
|
|
13
|
Accelerating evolutionary algorithms with gaussian process fitness function models
– Büche, Schraudolph, et al.
- 2005
|
|
12
|
Managing approximate models in evolutionary aerodynamic design optimization
– Jin, Olhofer, et al.
- 2001
|
|
7
|
Metamodel-assisted multiobjective optimisation strategies and their application in airfoil design
– Emmerich, Naujoks
- 2004
|
|
6
|
Evolution strategies assisted by Gaussian processes with improved pre-selection criterion
– Ulmer, Streichert, et al.
|
|
5
|
Metamodel-assisted SMS-EMOA applied to airfoil optimization tasks
– Naujoks, Beume, et al.
- 2005
|
|
4
|
A Reduced-Cost Multi-Objective Optimization Method based on the Pareto Front Technique
– Giotis, Giannakoglou, et al.
- 2000
|
|
3
|
Metamodel-assisted optimisation with constraints: A case study in material process design
– Emmerich, Jakumeit
- 2003
|
|
3
|
Low cost stochastic optimisation for engineering applications
– Bäck
- 2001
|
|
2
|
Forging process optimization
– Do, Fourment
|
|
2
|
Sensitivity analysis and optimization algorithms for 3d forging process design
– Do, Fourment, et al.
- 2004
|
|
2
|
Optimization of a gas turbine blade casting using evolution strategies and kriging
– Emmerich, Jakumeit
|
|
2
|
Einsatz von metamodell-gestützten Evolutionsstrategien in der elektromagnetischen Feldoptimierung
– Varcol
- 2003
|
