Normal-bundle bootstrap

DOI10.1137/20M1356002MaRDI QIDQ4999394zbMATH OpenOpenAlexFDO

Authors Ruda Zhang, R. G. Ghanem

Publication date 6 July 2021

Published in SIAM Journal on Mathematics of Data Science (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2007.13869

zbMATH Keywords

data augmentationresamplingprobabilistic learningdata manifold

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05) Nonparametric statistical resampling methods (62G09) Statistics on manifolds (62R30) Higher-dimensional and -codimensional surfaces in Euclidean and related (n)-spaces (53A07) Computational methods for attractors of dynamical systems (37M22)

Abstract: Probabilistic models of data sets often exhibit salient geometric structure. Such a phenomenon is summed up in the manifold distribution hypothesis, and can be exploited in probabilistic learning. Here we present normal-bundle bootstrap (NBB), a method that generates new data which preserve the geometric structure of a given data set. Inspired by algorithms for manifold learning and concepts in differential geometry, our method decomposes the underlying probability measure into a marginalized measure on a learned data manifold and conditional measures on the normal spaces. The algorithm estimates the data manifold as a density ridge, and constructs new data by bootstrapping projection vectors and adding them to the ridge. We apply our method to the inference of density ridge and related statistics, and data augmentation to reduce overfitting.

Recommendations

• Bootstrap-based normal reconstruction
• Multilayer bootstrap networks
• Bootstrapping descriptors for non-Euclidean data
• Bootstrapping complex functions
• Testing the manifold hypothesis

Cites work

• scientific article; zbMATH DE number 3729119 (Why is no real title available?)
• scientific article; zbMATH DE number 708500 (Why is no real title available?)
• Asymptotic theory for density ridges
• Convergence of the Iterates of Descent Methods for Analytic Cost Functions
• Data-driven probability concentration and sampling on manifold
• Deep learning
• Differential Topology
• Differential geometry and mathematical physics. I: Manifolds, Lie groups and Hamiltonian systems
• Geodesic Monte Carlo on embedded manifolds
• Hybrid Monte Carlo methods for sampling probability measures on submanifolds
• Introduction to Smooth Manifolds
• Jackknife, bootstrap and other resampling methods in regression analysis
• Locally defined principal curves and surfaces
• Manifold estimation and singular deconvolution under Hausdorff loss
• Monte Carlo on Manifolds: Sampling Densities and Integrating Functions
• Nonlinear oscillations, dynamical systems, and bifurcations of vector fields
• Nonparametric inference on manifolds. With applications to shape spaces
• Nonparametric modal regression
• Nonparametric ridge estimation
• Nonparametric statistics on manifolds and their applications to object data analysis
• On Multiple Eigenvalues of Matrices Depending on Several Parameters
• Pattern recognition and machine learning.
• Principal Curves
• Sampling from a Manifold
• Sampling of Bayesian posteriors with a non-Gaussian probabilistic learning on manifolds from a small dataset
• Statistics on special manifolds
• The geometric foundations of Hamiltonian Monte Carlo

Cited in

(2)

• Data driven Dirichlet sampling on manifolds
• Gaussian Process Subspace Prediction for Model Reduction

Describes a project that uses

Uses Software

• PRMLT
• bootstrap

This page was built for publication: Normal-bundle bootstrap

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4999394)