Trust-region Method for Derivative Free Case
No Thumbnail Available
Date
2015-11
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
III this project a derivative-free (DF) surrogate-based trust regioll
optimization approach is proposed. In the proposed approach, quadratic
surrogate models are constructed and successively IIpdated. The generated
surrogate model is then optimized instead oj the underlined objective
j unction over trust regions. Trullcated conjugate gradients are employed to
find the optimal point within each trust region. The approach COl/structs the
initial quadratic surrogate model using few data points oj order 0(11), where
II is the lIumber oj design variables. The proposed approach adopts
weighted least squares fitting Jar updatillg the SlIrrogate model illstead oj
interpolation which ;s commollly used ill DF optimization, This makes the
approach more suitable/or s tochastic optim ization and for funcliolls subject
[0 I/Il1llerical error. The weights are assigl/ed 10 give more emphasis to
points close to th e current center point.
Description
Keywords
• Derivative-free optimization; • Quadratic surrogate lIlodel; • Trust region;