Global versus local search in constrained optimization of computer models

By Matthias Schonlau, William J. Welch, Donald R. Jones. 1998

How researchers see this paper

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

Citations in the corpus

An evaluation of adaptive surrogate modeling based optimization with two benchmark problems (2014, methodology): “For this reason, MEI tends to put more points near the boundary of the feasible space in the initial search stage.”
Taking the Human Out of the Loop: A Review of Bayesian Optimization (2016, methodology): “The term Bayesian optimization was coined in the 1970s [115], but a popular version of the method has been known as efficient global optimization in the experimental design literature since the 1990s [134].”
Recent Advances in Bayesian Optimization (2022, background): “Interestingly, similar ideas have been discussed in [160] and revisited in [51].”
A Taxonomy of Global Optimization Methods Based on Response Surfaces (2001, methodology): “In the literature, this method has been pursued by Schonlau et al. (1997), and by Sasena (2000).”
Quantile-Based Optimization of Noisy Computer Experiments With Tunable Precision (2013, background): “The parameter β tunes the level of reliability wanted on the final result (which plays a similar role as the power parameter of the generalized improvement of Schonlau, Welch and Jones ( 1998 )).”
Mode-pursuing sampling method for global optimization on expensive black-box functions (2004, methodology): “[5, 6], where the authors apply a Bayesian method to estimate a kriging model, and to gradually identify points in the space to update the model and perform the optimization.”
A new acquisition function for Bayesian optimization based on the moment-generating function (2017, background): “This means the GEI functions are, in fact, indifferent when the parameter g increases.”
A Tutorial on Bayesian Optimization (2018, background): “This was proposed in Section 4 of Schonlau et al. (1998) and studied independently by Gardner et al. (2014).”
Exponential Regret Bounds for Gaussian Process Bandits with Deterministic Observations (2012, background): “We refer the reader to (Močkus, 1982; Schonlau et al., 1998; Gramacy et al., 2004; Brochu et al., 2007; Lizotte, 2008; Martinez–Cantin et al., 2009; Garnett et al., 2010) for practical examples.”
Cooling Strategies for the Moment-Generating Function in Bayesian Global Optimization (2018, background): “Generalized Expected Improvement (GEI) [17] realizes this idea by introducing a free parameter g as the order of the moment: GEI(x; g) = E{I(x)g ∣ Y } , g = 1,2, . . . (3) The larger the value of g, the more explorative the algorithm will perform and vice versa.”
Auto-WEKA: combined selection and hyperparameter optimization of classification algorithms (2012, methodology): “We use k-fold cross-validation [19], which splits the training data into k equal-sized partitions D(1)valid, . . . ,D (k) valid, and sets D(i)train = D \ D (i) valid for i = 1, . . . , k. 1”
Constrained Bayesian Optimization with Noisy Experiments (2017, background): “Schonlau et al. (1998) deﬁne the improvement of a candidate x over f ∗ c to be 0 if x is infeasible, and otherwise to be the usual improvement.”
Sequential parameter optimization (2005, background): “Computing the candidates model value, the generalized expected improvement criterion is determined next [4].”
Survey of modeling and optimization strategies to solve high-dimensional design problems with computationally-expensive black-box functions (2010, background): “…for computationally-expensive black-box functions It can be seen from literature that implementation of optimization of computationally expensive black-box functions often uses a cheap or approximate model as a surrogate of the expensive model (e.g. Jones et al. 1998; Schonlau et al. 1998).”
Sequential design of computer experiments to minimize integrated response functions (2000, background): “Jones, Schonlau and Welch (1998) and Schonlau, Welch and Jones (1998) introduced a criterion-based sequential strategy for response minimization.”
Learning manipulation skills from a single demonstration (2018, methodology): “There exist different methods of including additional constraints in Bayesian optimization (Gardner et al., 2014; Gelbart et al., 2014; Gramacy and Lee, 2011; Schonlau et al., 1998).”
BoTorch: Programmable Bayesian Optimization in PyTorch (2019, methodology): “[83] Matthias Schonlau, William J.”
Democratizing Data Science through Interactive Curation of ML Pipelines (2019, methodology): “In our work we use the widely adopted expected improvement (EI) [31] as sampling policy, due its ability to balance exploration (search in unexplored regions) and exploitation (search in promising regions) [16, 31, 38].”