Optimal robust expensive optimization is tractable

Philippe Rolet 1, 2 Michèle Sebag 1, 2 Olivier Teytaud 1, 2
2 TAO - Machine Learning and Optimisation
LRI - Laboratoire de Recherche en Informatique, UP11 - Université Paris-Sud - Paris 11, Inria Saclay - Ile de France, CNRS - Centre National de la Recherche Scientifique : UMR8623
Abstract : Following a number of recent papers investigating the possibility of optimal comparison-based optimization algorithms for a given distribution of probability on fitness functions, we (i) discuss the comparison-based constraints (ii) choose a setting in which theoretical tight bounds are known (iii) develop a careful implementation using billiard algorithms, Upper Confidence trees and (iv) experimentally test the tractability of the approach. The results, on still very simple cases, show that the approach, yet still preliminary, could be tested successfully until dimension 10 and horizon 50 iterations within a few hours on a standard computer, with convergence rate far better than the best algorithms.
Document type :
Conference papers
Complete list of metadatas

Cited literature [24 references]  Display  Hide  Download

https://hal.inria.fr/inria-00374910
Contributor : Olivier Teytaud <>
Submitted on : Friday, April 10, 2009 - 11:38:16 AM
Last modification on : Thursday, April 5, 2018 - 12:30:12 PM
Long-term archiving on : Thursday, June 10, 2010 - 8:17:47 PM

File

balopt.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : inria-00374910, version 1

Collections

Citation

Philippe Rolet, Michèle Sebag, Olivier Teytaud. Optimal robust expensive optimization is tractable. Gecco 2009, ACM, 2009, Montréal, Canada. 8 p. ⟨inria-00374910⟩

Share

Metrics

Record views

430

Files downloads

324