Skip to Main content Skip to Navigation
Conference papers

Comparison of NEWUOA with Different Numbers of Interpolation Points on the BBOB Noiseless Testbed

Raymond Ros 1
1 TAO - Machine Learning and Optimisation
CNRS - Centre National de la Recherche Scientifique : UMR8623, Inria Saclay - Ile de France, UP11 - Université Paris-Sud - Paris 11, LRI - Laboratoire de Recherche en Informatique
Abstract : In this paper, we study the performances of the NEW Unconstrained Optimization Algorithm (NEWUOA) with different numbers of interpolation points. NEWUOA is a trust region method, the number of points used to build the surrogate model is an input parameter of the algorithm. We compare the performances of NEWUOA using three different number of points in search spaces of dimension from two to forty on problems from the BBOB 2009 noiseless function testbed. In particular we study the performances of an `average' number of interpolation points that scales like the dimension of the search space to the power 3/2. Using this number of interpolation points is expectedly faster than using the maximum number of interpolation points (scaling like the square of the dimension), though it does not grant better performances than using a number of interpolation points scaling like the dimension.
Document type :
Conference papers
Complete list of metadatas

https://hal.inria.fr/inria-00473774
Contributor : Raymond Ros <>
Submitted on : Friday, April 16, 2010 - 2:00:39 PM
Last modification on : Wednesday, September 16, 2020 - 5:06:55 PM
Long-term archiving on: : Tuesday, September 28, 2010 - 12:40:42 PM

Files

ws1p10-ros.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : inria-00473774, version 1

Collections

Citation

Raymond Ros. Comparison of NEWUOA with Different Numbers of Interpolation Points on the BBOB Noiseless Testbed. Genetic and Evolutionary Computation Conference 2010, Jul 2010, Portland, OR, United States. ⟨inria-00473774⟩

Share

Metrics

Record views

212

Files downloads

502