Toward an optimal ensemble of kernel-based approximations with engineering applications

Sanchez, E. J.; Pintos, Salvador; Queipo, Néstor V.

doi:10.1007/s00158-007-0159-6

Cited by 80 publications

(16 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The main idea behind model combination is to construct a combined model that yields better prediction accuracy than any single model. Usually one of the following two problems is tackled: building an ensemble of one particular model type with different hyperparameter settings [20] or building an ensemble of heterogeneous models with fixed hyperparameter settings [18,13,24]. Hyperparameters are parameters of model or learner.…”

Section: Model Combinationmentioning

confidence: 99%

Comparison of Classical and Sequential Design of Experiments in Note Onset Detection

Bauer

Schiffner

Weihs

2013

Algorithms From and for Nature and Life

View full text Add to dashboard Cite

Design of experiments is an established approach to parameter optimization of industrial processes. In many computer applications however it is usual to optimize the parameters via genetic algorithms. The main idea of this work is to apply design of experiment's techniques to the optimization of computer processes. The major problem here is finding a compromise between model validity and costs, which increase with the number of experiments. The second relevant problem is choosing an appropriate model, which describes the relationship between parameters and target values. One of the recent approaches here is model combination, which can be used in sequential designs in order to improve automatic prediction of the next trial point.In this paper a musical note onset detection algorithm will be optimized using sequential parameter optimization with model combination. It will be shown that parameter optimization via design of experiments leads to better values of the target variable than usual parameter optimization via grid search or genetic optimization algorithms. Furthermore, the results of this application study reveal, whether the combination of many models brings improvements in finding the optimal parameter setting.

show abstract

Section: Model Combinationmentioning

confidence: 99%

Comparison of Classical and Sequential Design of Experiments in Note Onset Detection

Bauer

Schiffner

Weihs

2013

Algorithms From and for Nature and Life

View full text Add to dashboard Cite

show abstract

“…Additionally, we also used the native neural networks MATLAB toolbox [45] and the code developed by Gunn [57] to run radial basis neural network and support vector regression algorithms, respectively. We use multiple instances of different surrogates in the same fashion of Viana et al [29] and Sanchez et al [58]. This is possible because kriging allows different instances by changing parameters such as basis and correlation functions.…”

Section: Appendix A: Cross Validation For Surrogate Selectionmentioning

confidence: 99%

Using Cross Validation to Design Conservative Surrogates

2010

View full text Add to dashboard Cite

The use of surrogates (also known as metamodels) for facilitating optimization and statistical analysis of computationally expensive simulations has become commonplace. Surrogate models are usually fit to be unbiased (i.e., the error expectation is zero). However, in certain applications, it might be important to safely estimate the response (e.g., in structural analysis, the maximum stress must not be underestimated in order to avoid failure). In this work we use safety margins to conservatively compensate for fitting errors associated with surrogates. We propose the use of cross validation for estimating the required safety margin for a desired level of conservativeness (percentage of safe predictions). The approach was tested on three algebraic examples for two basic surrogates: namely, kriging and polynomial response surface. For these examples we found that cross validation is effective for selecting the safety margin. We also applied the approach to the probabilistic design optimization of a composite laminate. This design under uncertainty example showed that the approach can be successfully used in engineering applications.

show abstract

“…Once a surrogate model has been chosen, there are many ways of introducing it into the original algorithm. Several approaches have been made in general surrogate-assisted evolutionary frameworks such as: integrating GAs with surrogate approximations (Queipo et al, 2005;Regis & Shoemaker, 2004) or landscape approximations (Knowles, 2006), the use of surrogate-guided evolutionary operators (Rasheed, 2002), surrogate-assisted local search (Lim et al, 2008;Wanner et al 2008), accelerating the optimization process using surrogate models, pre-selection approaches (Giannakoglou, 2002;Praveen & Duvigneau, 2009), multiple surrogates (Acar & Rais-Rohani, 2008;Lim et al, 2008;Sanchez et al 2007), and coevolution of fitness predictors (Schmidt & Lipson, 2008). However, no Surrogate-Assisted AIS algorithm seems to have been proposed in the literature so far.…”

Section: Surrogate-assisted Artificial Immune Systemmentioning

confidence: 99%

Surrogate-Assisted Artificial Immune Systems for Expensive Optimization Problems

Bernardino¹,

Goliatt²,

Barbosa³

et al. 2009

Evolutionary Computation

View full text Add to dashboard Cite

Toward an optimal ensemble of kernel-based approximations with engineering applications

Cited by 80 publications

References 31 publications

Comparison of Classical and Sequential Design of Experiments in Note Onset Detection

Comparison of Classical and Sequential Design of Experiments in Note Onset Detection

Using Cross Validation to Design Conservative Surrogates

Surrogate-Assisted Artificial Immune Systems for Expensive Optimization Problems

Contact Info

Product

Resources

About