A supermartingale approach to Gaussian process based sequential design of experiments
From MaRDI portal
Publication:2325344
DOI10.3150/18-BEJ1074zbMATH Open1428.62369arXiv1608.01118OpenAlexW2489204415MaRDI QIDQ2325344FDOQ2325344
Authors: Julien Bect, François Bachoc, D. Ginsbourger 
Publication date: 25 September 2019
Published in: Bernoulli (Search for Journal in Brave)
Abstract: Gaussian process (GP) models have become a well-established frameworkfor the adaptive design of costly experiments, and notably of computerexperiments. GP-based sequential designs have been found practicallyefficient for various objectives, such as global optimization(estimating the global maximum or maximizer(s) of a function),reliability analysis (estimating a probability of failure) or theestimation of level sets and excursion sets. In this paper, we studythe consistency of an important class of sequential designs, known asstepwise uncertainty reduction (SUR) strategies. Our approach relieson the key observation that the sequence of residual uncertaintymeasures, in SUR strategies, is generally a supermartingale withrespect to the filtration generated by the observations. Thisobservation enables us to establish generic consistency results for abroad class of SUR strategies. The consistency of several popularsequential design strategies is then obtained by means of this generalresult. Notably, we establish the consistency of two SUR strategiesproposed by Bect, Ginsbourger, Li, Picheny and Vazquez (Stat. Comp.,2012)---to the best of our knowledge, these are the first proofs ofconsistency for GP-based sequential design algorithms dedicated to theestimation of excursion sets and their measure. We also establish anew, more general proof of consistency for the expected improvementalgorithm for global optimization which, unlike previous results inthe literature, applies to any GP with continuous sample paths.
Full work available at URL: https://arxiv.org/abs/1608.01118
Recommendations
- Advances in the sequential design of computer experiments based on active learning
- Sequential design of computer experiments for the estimation of a probability of failure
- Convergence rates of efficient global optimization algorithms
- Gaussian process surrogates for failure detection: a Bayesian experimental design approach
- scientific article; zbMATH DE number 1556153
convergencestepwise uncertainty reductionactive learningsequential design of experimentssupermartingaleuncertainty functional
Learning and adaptive systems in artificial intelligence (68T05) Gaussian processes (60G15) Sequential statistical design (62L05) Generalizations of martingales (60G48)
Cites Work
- Efficient global optimization of expensive black-box functions
- Title not available (Why is that?)
- Interpolation of spatial data. Some theory for kriging
- The design and analysis of computer experiments.
- Design and analysis of computer experiments. With comments and a rejoinder by the authors
- Information-Theoretic Regret Bounds for Gaussian Process Optimization in the Bandit Setting
- Title not available (Why is that?)
- Title not available (Why is that?)
- Title not available (Why is that?)
- Probabilistic numerics and uncertainty in computations
- Theory of Random Sets
- Sequential design of computer experiments for the estimation of a probability of failure
- An informational approach to the global optimization of expensive-to-evaluate functions
- Convergence properties of the expected improvement algorithm with fixed mean and covariance functions
- Statistics for spatial data
- Probability in Banach spaces. Isoperimetry and processes
- Stochastic processes with sample paths in reproducing kernel Hilbert spaces
- Title not available (Why is that?)
- Functional analysis, Sobolev spaces and partial differential equations
- Changes in utility as information
- Bayes-Hermite quadrature
- Level Sets and Extrema of Random Processes and Fields
- Title not available (Why is that?)
- Title not available (Why is that?)
- Average-case analysis of numerical problems
- Title not available (Why is that?)
- Convergence rates of efficient global optimization algorithms
- Learning functions and approximate Bayesian computation design: ABCD
- The knowledge-gradient policy for correlated normal beliefs
- Title not available (Why is that?)
- Uncertainty, Information, and Sequential Experiments
- On degeneracy and invariances of random fields paths with applications in Gaussian process modelling
- Probability theory. An analytic view.
- A Knowledge-Gradient Policy for Sequential Information Collection
- Pushing the Limits of Contemporary Statistics: Contributions in Honor of Jayanta K. Ghosh
- Title not available (Why is that?)
- A Bayesian approach to constrained single- and multi-objective optimization
- Bayesian adaptive reconstruction of profile optima and optimizers
- The correlated knowledge gradient for simulation optimization of continuous parameters using Gaussian process regression
- Jensen's inequality for a convex vector-valued function on an infinite-dimensional space
- Estimating functions evaluated by simulation: a Bayesian/analytic approach
- A note on the measurability of the pseudo-inverse
- Univariate interpolation by exponential functions and Gaussian RBFs for generic sets of nodes
- Bayesian Subset Simulation
- Examples of inconsistency in optimization by expected improvement
- Gaussian process surrogates for failure detection: a Bayesian experimental design approach
Cited In (27)
- Bayesian quadrature, energy minimization, and space-filling design
- Sequential Design of Multi-Fidelity Computer Experiments: Maximizing the Rate of Stepwise Uncertainty Reduction
- Locally induced Gaussian processes for large-scale simulation experiments
- Uncertainty Quantification and Experimental Design for Large-Scale Linear Inverse Problems under Gaussian Process Priors
- Maximum likelihood estimation for Gaussian processes under inequality constraints
- Advances in the sequential design of computer experiments based on active learning
- Surrogate-based sequential Bayesian experimental design using non-stationary Gaussian processes
- Parallel Gaussian process surrogate Bayesian inference with noisy likelihood evaluations
- Learning excursion sets of vector-valued Gaussian random fields for autonomous ocean sampling
- Expected improvement for expensive optimization: a review
- Gaussian process optimization with failures: classification and convergence proof
- Sequential construction and dimension reduction of Gaussian processes under inequality constraints
- Technical note—Knowledge gradient for selection with covariates: Consistency and computation
- Adaptive design for Gaussian process regression under censoring
- Adaptive Design of Experiments for Conservative Estimation of Excursion Sets
- Moderate deviations inequalities for Gaussian process regression
- Disintegration of Gaussian measures for sequential assimilation of linear operator data
- Title not available (Why is that?)
- A Bayesian optimization approach to find Nash equilibria
- Evaluating Gaussian process metamodels and sequential designs for noisy level set estimation
- Sequential Bayesian optimal experimental design for structural reliability analysis
- Finite-Dimensional Gaussian Approximation with Linear Inequality Constraints
- Data-driven stochastic inversion via functional quantization
- Bayesian Optimization with Expensive Integrands
- Adaptive spatial designs minimizing the integrated Bernoulli variance in spatial logistic regression models -- with an application to benthic habitat mapping
- Composite likelihood estimation for a Gaussian process under fixed domain asymptotics
- Computational enhancements to Bayesian design of experiments using Gaussian processes
Uses Software
This page was built for publication: A supermartingale approach to Gaussian process based sequential design of experiments
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2325344)
