Nonlinear dimension reduction for surrogate modeling using gradient information

DOI10.1093/IMAIAI/IAAC006MaRDI QIDQ5878248zbMATH OpenOpenAlexFDO

Authors Daniele Bigoni, Youssef M. Marzouk, Clémentine Prieur, Olivier Zahm

Publication date 20 February 2023

Published in Information and Inference: A Journal of the IMA (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2102.10351

zbMATH Keywords

nonlinear dimension reductionhigh-dimensional approximationfeature mapPoincaré inequalityadaptive polynomial approximation

Mathematics Subject Classification ID

Multidimensional problems (41A63) Approximation by polynomials (41A10) Algorithms for approximation of functions (65D15) Numerical approximation of high-dimensional functions; sparse grids (65D40)

Abstract: We introduce a method for the nonlinear dimension reduction of a high-dimensional function $u:mathbb{R}^{d}ightarrowmathbbR$, $dgg1$. Our objective is to identify a nonlinear feature map $g:mathbb{R}^{d}ightarrowmathbb{R}^m$, with a prescribed intermediate dimension $mlld$, so that $u$ can be well approximated by $fcircg$ for some profile function $f:mathbb{R}^{m}ightarrowmathbbR$. We propose to build the feature map by aligning the Jacobian $ablag$ with the gradient $ablau$, and we theoretically analyze the properties of the resulting $g$. Once $g$ is built, we construct $f$ by solving a gradient-enhanced least squares problem. Our practical algorithm makes use of a sample ${x^{(i)},u(x^{(i)}),ablau(x^{(i)})}_{i=1}^N$ and builds both $g$ and $f$ on adaptive downward-closed polynomial spaces, using cross validation to avoid overfitting. We numerically evaluate the performance of our algorithm across different benchmarks, and explore the impact of the intermediate dimension $m$. We show that building a nonlinear feature map $g$ can permit more accurate approximation of $u$ than a linear $g$, for the same input data set.

Recommendations

• Symmetry results for decay solutions of elliptic systems in the whole space
• Nonlinear level set learning for function approximation on sparse data with applications to parametric differential equations
• Gradient-based dimension reduction of multivariate vector-valued functions
• High-dimensional nonlinear multi-fidelity model with gradient-free active subspace method
• Gradient-enhanced surrogate modeling based on proper orthogonal decomposition

Cited in

(16)

• Level Set Learning with Pseudoreversible Neural Networks for Nonlinear Dimension Reduction in Function Approximation
• On the representation and learning of monotone triangular transport maps
• Principal feature detection via -Sobolev inequalities
• Data-driven surrogates for high dimensional models using Gaussian process regression on the Grassmann manifold
• Model Reduction for Nonlinear Systems by Balanced Truncation of State and Gradient Covariance
• Gradient-enhanced surrogate modeling based on proper orthogonal decomposition
• A novel and fully automated coordinate system transformation scheme for near optimal surrogate construction
• Surrogate to Poincaré inequalities on manifolds for dimension reduction in nonlinear feature spaces
• Nonlinear level set learning for function approximation on sparse data with applications to parametric differential equations
• High-dimensional nonlinear multi-fidelity model with gradient-free active subspace method
• Prior normalization for certified likelihood-informed subspace detection of Bayesian inverse problems
• Sharp detection of low-dimensional structure in probability measures via dimensional logarithmic Sobolev inequalities
• Diffeomorphism-based feature learning using Poincaré inequalities on augmented input space
• An adaptive learning strategy for surrogate modeling of high-dimensional functions - application to unsteady hypersonic flows in chemical nonequilibrium
• A review of recent advances in surrogate models for uncertainty quantification of high-dimensional engineering applications
• Conditional Karhunen-Loève regression model with basis adaptation for high-dimensional problems: uncertainty quantification and inverse modeling

This page was built for publication: Nonlinear dimension reduction for surrogate modeling using gradient information

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5878248)