In libKriging/dolka: Utility functions for Bayesian optimization

This document is intended to be read from the generated html output TODO.html. However wherever this output is not suitably rendered you can read the mardown version TODO.md which seems correct up to some problems for math.

General

• The one-point optimization criteria such as \code{"EI"} and their derivatives could optionnaly be computed by the \code{predict} method or by a new method say \code{optCrit1}.

Existing functions

• contours Allow to pass further arguments to predict or to the functions other and otherGrad. For instance we may want to change type in predict or proxy in EI used as other.

• [x] genoud_cache Check that all arguments of fn are well used. Of course, optimization criteria such as EI needs model.

New functions

• Add a optim_cache function to use optim without having to manually cache the gradient in an environment.

libKriging/dolka documentation built on April 14, 2022, 7:17 a.m.