Advanced search
2 files | 10.32 MB

Pareto-based multi-output metamodeling with active learning

Dirk Gorissen (UGent) , Ivo Couckuyt (UGent) , Eric Laermans (UGent) and Tom Dhaene (UGent)
Author
Organization
Abstract
When dealing with computationally expensive simulation codes or process measurement data, global surrogate modeling methods are firmly established as facilitators for design space exploration, sensitivity analysis, visualization and optimization. Popular surrogate model types include neural networks, support vector machines, and splines. In addition, the cost of each simulation mandates the use of active learning strategies where data points (simulations) are selected intelligently and incrementally. When applying surrogate models to multi-output systems, the hyperparameter optimization problem is typically formulated in a single objective way. The different response outputs are modeled separately by independent models. Instead, a multi-objective approach would benefit the domain expert by giving information about output correlation, facilitate the generation of diverse ensembles, and enable automatic model type selection for each output on the fly. This paper outlines a multi-objective approach to surrogate model generation including its application to two problems.
Keywords
MULTIOBJECTIVE OPTIMIZATION, NETWORKS

Downloads

  • (...).pdf
    • full text
    • |
    • UGent only
    • |
    • PDF
    • |
    • 7.62 MB
  • (...).PDF
    • full text
    • |
    • UGent only
    • |
    • PDF
    • |
    • 2.71 MB

Citation

Please use this url to cite or link to this publication:

Chicago
Gorissen, Dirk, Ivo Couckuyt, Eric Laermans, and Tom Dhaene. 2009. “Pareto-based Multi-output Metamodeling with Active Learning.” In Communications in Computer and Information Science, ed. Dominic Palmer-Brown, Chrisina Draganova, Elias Pimenidis, and Haris Mouratidis, 43:389–400. Berlin, Germany: Springer.
APA
Gorissen, D., Couckuyt, I., Laermans, E., & Dhaene, T. (2009). Pareto-based multi-output metamodeling with active learning. In D. Palmer-Brown, C. Draganova, E. Pimenidis, & H. Mouratidis (Eds.), COMMUNICATIONS IN COMPUTER AND INFORMATION SCIENCE (Vol. 43, pp. 389–400). Presented at the 11th International Conference on Engineering Applications of Neural Networks (EANN 2009), Berlin, Germany: Springer.
Vancouver
1.
Gorissen D, Couckuyt I, Laermans E, Dhaene T. Pareto-based multi-output metamodeling with active learning. In: Palmer-Brown D, Draganova C, Pimenidis E, Mouratidis H, editors. COMMUNICATIONS IN COMPUTER AND INFORMATION SCIENCE. Berlin, Germany: Springer; 2009. p. 389–400.
MLA
Gorissen, Dirk, Ivo Couckuyt, Eric Laermans, et al. “Pareto-based Multi-output Metamodeling with Active Learning.” Communications in Computer and Information Science. Ed. Dominic Palmer-Brown et al. Vol. 43. Berlin, Germany: Springer, 2009. 389–400. Print.
@inproceedings{803256,
  abstract     = {When dealing with computationally expensive simulation codes or process measurement data, global surrogate modeling methods are firmly established as facilitators for design space exploration, sensitivity analysis, visualization and optimization. Popular surrogate model types include neural networks, support vector machines, and splines. In addition, the cost of each simulation mandates the use of active learning strategies where data points (simulations) are selected intelligently and incrementally. When applying surrogate models to multi-output systems, the hyperparameter optimization problem is typically formulated in a single objective way. The different response outputs are modeled separately by independent models. Instead, a multi-objective approach would benefit the domain expert by giving information about output correlation, facilitate the generation of diverse ensembles, and enable automatic model type selection for each output on the fly. This paper outlines a multi-objective approach to surrogate model generation including its application to two problems.},
  author       = {Gorissen, Dirk and Couckuyt, Ivo and Laermans, Eric and Dhaene, Tom},
  booktitle    = {COMMUNICATIONS IN COMPUTER AND INFORMATION SCIENCE},
  editor       = {Palmer-Brown, Dominic and Draganova, Chrisina and Pimenidis, Elias and Mouratidis, Haris},
  isbn         = {9783642039690},
  language     = {eng},
  location     = {London, UK},
  pages        = {389--400},
  publisher    = {Springer},
  title        = {Pareto-based multi-output metamodeling with active learning},
  url          = {http://dx.doi.org/10.1007/978-3-642-03969-0\_36},
  volume       = {43},
  year         = {2009},
}

Altmetric
View in Altmetric
Web of Science
Times cited: