Stats, Optimization, and Machine Learning Seminar - Nathaniel Mathews
discussion of paper "Constrained Global Optimization of Expensive Black Box Functions Using Radial Basis Functions"
Ìý
The authors (Regis et al.) propose an iterative response-surface model for optimization which is well suited to nonlinear constraints on nonconvex objectives, and is meant to allow relatively fast optimization for high-dimensional problems. We will discuss this paper.
References:
"Constrained Global Optimization of Expensive Black Box Functions Using Radial Basis Functions" by Regis, Rommel G; Shoemaker, Christine A. Journal of Global Optimization; Dordrecht Vol. 31, Iss. 1,Ìý (Jan 2005): 153-171.Ìý
For a larger context, see Gutmann 2001Ìý
And an extension, by the same author: Regis 2011:Ìý