Examples of inconsistency in optimization by expected improvement
From MaRDI portal
Abstract: We consider the 1D Expected Improvement optimization based on Gaussian processes having spectral densities converging to zero faster than exponentially. We give examples of problems where the optimization trajectory is not dense in the design space. In particular, we prove that for Gaussian kernels there exist smooth objective functions for which the optimization does not converge on the optimum.
Recommendations
- Convergence rates of efficient global optimization algorithms
- Convergence properties of the expected improvement algorithm with fixed mean and covariance functions
- On the convergence rates of expected improvement methods
- Expected improvement for expensive optimization: a review
- A complete expected improvement criterion for Gaussian process assisted highly constrained expensive optimization
Cites work
- A radial basis function method for global optimization
- A taxonomy of global optimization methods based on response surfaces
- Bayesian algorithms for one-dimensional global optimization
- Convergence properties of the expected improvement algorithm with fixed mean and covariance functions
- Convergence rates of efficient global optimization algorithms
- Efficient global optimization of expensive black-box functions
- Gaussian processes for machine learning.
- Global optimization
- scientific article; zbMATH DE number 996424 (Why is no real title available?)
- scientific article; zbMATH DE number 43985 (Why is no real title available?)
- scientific article; zbMATH DE number 3612796 (Why is no real title available?)
- scientific article; zbMATH DE number 2107175 (Why is no real title available?)
Cited in
(7)- Expected improvement for expensive optimization: a review
- A supermartingale approach to Gaussian process based sequential design of experiments
- On the convergence rates of expected improvement methods
- Asymptotic Bounds for Smoothness Parameter Estimates in Gaussian Process Interpolation
- Simple-regret rates and minimax optimality of fixed-prior expected improvement in Matérn and squared-exponential RKHSs
- Approximation in Hilbert spaces of the Gaussian and related analytic kernels
- Convergence properties of the expected improvement algorithm with fixed mean and covariance functions
This page was built for publication: Examples of inconsistency in optimization by expected improvement
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2393071)
