A new sequential optimal sampling method for radial basis functions

摘要

The sampling procedure is crucial for accuracy of model approximation for a given metamodel. In this paper, we propose a new sequential sampling method for radial basis functions. At each sampling step, we provided a criterion to determine the optimal sampling point, which maximizes the value of the product of curvature and square of minimum distance to other design sites. Experiments proved that using the same sampling size the highest accuracy and efficiency can be obtained to approximate an objective function under this criterion.