01.1 - Aircraft Design and Integrated System (Basics and Theory)
EXPLORATION OF EFFICIENT HYPERPARAMETERS ADAPTION OF SUPPORT VECTOR REGRESSION FOR AERODYNAMIC DESIGN
K.-S. Zhang¹, H.-L. Qiao¹, P.-H. Wang¹, Y.-Q. Du¹, Z.-H. Han¹; ¹School of Aeronautics, Northwestern Polytechnical University, China
This work aims to explore the hyperparameters adaption strategy for efficient support vector regression as it critically affects modeling accuracy and computational cost. The objective of hyperparameter adaption is estimated by cross-validation or leave-one-out bound method with smaller computational cost. The hyperparameters are optimized by a global optimization algorithm, such as Genetic Algorithm, Bayesian Optimization or Covariance Matrix Adaptation Evolution Strategy. The different strategies are compared via series of numerical examples and aerodynamic design cases.rn