A benchmark of kriging-based infill criteria for noisy optimization

By Victor Picheny, Tobias Wagner, David Ginsbourger. Structural And Multidisciplinary Optimization, 2013

403 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Hyperparameter optimization: Foundations, algorithms, best practices, and open challenges (2021, background): “This can either be the best observed λ during optimization, the best (mean, or quantile) predicted λ from the archive according to the surrogate model (Jalali et al., 2017; Picheny et al., 2013), or the best predicted λ overall (Scott et al., 2011).”
Recent Advances in Bayesian Optimization (2022, background): “A direct approach is to replace f (x) by some sensible values, which is called expected improvement with łplug-in" [145].”
Gaussian Processes for history-matching: application to an unconventional gas reservoir (2017, background): “[135, 136]), in our experience for many optimization and history-matching problems, the Matérn class with ν = 3/2 provides good results with favorable convergence rate.”
Bayesian tuning and bandits : an extensible, open source library for AutoML (2018, background): “com/ 2A Rosenbrock function has a hard-to-find global minima residing in an easy-to-find valley [23].”
Student-t Processes as Alternatives to Gaussian Processes (2014, background): “Hartmann function This is a function with 6 local minima in [0, 1]6 [Picheny et al., 2013].”
Constrained Bayesian Optimization with Noisy Experiments (2017, methodology): “Picheny et al. (2013) extend this idea by substituting a GP quantile in the place of the mean, and then optimize expected improvement of that quantile.”