Fast Computation of the Multi-Points Expected Improvement with Applications in Batch Selection. Chevalier, C. & Ginsbourger, D. In Nicosia, G. & Pardalos, P., editors, Learning and Intelligent Optimization, pages 59–69, Berlin, Heidelberg, 2013. Springer Berlin Heidelberg.
abstract   bibtex   
The Multi-points Expected Improvement criterion (or $}{$q$}{$-EI) has recently been studied in batch-sequential Bayesian Optimization. This paper deals with a new way of computing $}{$q$}{$-EI, without using Monte-Carlo simulations, through a closed-form formula. The latter allows a very fast computation of $}{$q$}{$-EI for reasonably low values of $}{$q$}{$(typically, less than 10). New parallel kriging-based optimization strategies, tested on different toy examples, show promising results.
@InProceedings{1Chevalier2013,
    author      = {Chevalier, Cl{\'e}ment and Ginsbourger, David},
    title       = {Fast Computation of the Multi-Points Expected Improvement with Applications in Batch Selection},
    address     = {Berlin, Heidelberg},
    booktitle   = {Learning and Intelligent Optimization},
    editor      = {Nicosia, Giuseppe and Pardalos, Panos},
    isbn        = {978-3-642-44973-4},
    pages       = {59--69},
    publisher   = {Springer Berlin Heidelberg},
    year        = {2013},
    abstract    = {The Multi-points Expected Improvement criterion (or {\$}{\$}q{\$}{\$}-EI) has recently been studied in batch-sequential Bayesian Optimization. This paper deals with a new way of computing {\$}{\$}q{\$}{\$}-EI, without using Monte-Carlo simulations, through a closed-form formula. The latter allows a very fast computation of {\$}{\$}q{\$}{\$}-EI for reasonably low values of {\$}{\$}q{\$}{\$}(typically, less than 10). New parallel kriging-based optimization strategies, tested on different toy examples, show promising results.}
}

Downloads: 0