This work explores methods in multi-fidelity and batch Bayesian Optimization, explaining why there is a close link between them and overviewing the current state-of-the-art. We propose a simple algorithm that allows practitioners to use any acquisition function well suited for their particular problem.
Link to research paper: https://arxiv.org/abs/2211.06149
Link to the code: https://github.com/jpfolch/MFBoom
JPF was funded by EPSRC through the Modern Statistics and Statistical Machine Learning (StatML) CDT (grant no. EP/S023151/1) and by BASF SE, Ludwigshafen am Rhein. The research was funded by Engineering & Physical Sciences Research Council (EPSRC) Fellowships to RM and CT (grant no. EP/P016871/1 and EP/T001577/1). CT acknowledges support from an Imperial College Research Fellowship. RM acknowledges support from the BASF / RAEng Research Chair in Data-Driven Optimization.
~
Connect with the Computational Optimisation Group at Imperial College London ...
Subscribe to the CogImperial YouTube channel: https://www.youtube.com/@cogimperial
Follow the presenter on Twitter: https://twitter.com/folch_pablo
Follow our research group on Twitter: https://twitter.com/CogImperial
Visit our official site for more information: https://optimisation.doc.ic.ac.uk
