06105760

a

06105760 Janusevskis, Janis Le Riche, Rodolphe Ginsbourger, David Girdziusas, Ramunas Expected improvements for the asynchronous parallel global optimization of expensive functions: potentials and challenges. Hamadi, Youssef (ed.) et al., Learning and intelligent optimization. 6th international conference, LION 6, Paris, France, January 16--20, 2012. Revised selected papers. Berlin: Springer (ISBN 978-3-642-34412-1/pbk). Lecture Notes in Computer Science 7219, 413-418 (2012). 2012 Berlin: Springer EN

doi:10.1007/978-3-642-34413-8_37

Summary: Sequential sampling strategies based on Gaussian processes are now widely used for the optimization of problems involving costly simulations. But Gaussian processes can also generate parallel optimization strategies. We focus here on a new, parameter free, parallel expected improvement criterion for asynchronous optimization. An estimation of the criterion, which mixes Monte Carlo sampling and analytical bounds, is proposed. Logarithmic speed-ups are measured on 1 and 9 dimensional functions.