An infinitesimal probabilistic model for principal component analysis of manifold valued data

DOI10.1007/S13171-018-0139-5zbMATH Open1426.62404arXiv1801.10341OpenAlexW2964325228WikidataQ129393026 ScholiaQ129393026MaRDI QIDQ2316992FDOQ2316992

Authors: Stefan Sommer 

Publication date: 7 August 2019

Published in: Sankhyā. Series A (Search for Journal in Brave)

Abstract: We provide a probabilistic and infinitesimal view of how the principal component analysis procedure (PCA) can be generalized to analysis of nonlinear manifold valued data. Starting with the probabilistic PCA interpretation of the Euclidean PCA procedure, we show how PCA can be generalized to manifolds in an intrinsic way that does not resort to linearization of the data space. The underlying probability model is constructed by mapping a Euclidean stochastic process to the manifold using stochastic development of Euclidean semimartingales. The construction uses a connection and bundles of covariant tensors to allow global transport of principal eigenvectors, and the model is thereby an example of how principal fiber bundles can be used to handle the lack of global coordinate system and orientations that characterizes manifold valued statistics. We show how curvature implies non-integrability of the equivalent of Euclidean principal subspaces, and how the stochastic flows provide an alternative to explicit construction of such subspaces. We describe estimation procedures for inference of parameters and prediction of principal components, and we give examples of properties of the model on embedded surfaces.

Full work available at URL: https://arxiv.org/abs/1801.10341

Recommendations

• Principal flows
• Scale and curvature effects in principal geodesic analysis
• Probabilistic geodesic models for regression and dimensionality reduction on Riemannian manifolds
• Statistical methods generalizing principal component analysis to non-Euclidean spaces
• Principal arc analysis on direct product manifolds

zbMATH Keywords

principal component analysisframe bundlestochastic developmentprobabilistic PCAanisotropic normal distributionsmanifold valued statistics

Mathematics Subject Classification ID

Factor analysis and principal components; correspondence analysis (62H25) Statistics on manifolds (62R30) Global differential geometry (53C99)

Cites Work

• Analysis of principal nested spheres
• Torus principal component analysis with applications to RNA structure
• Title not available (Why is that?)
• Stochastic analysis on manifolds
• Probabilistic Principal Component Analysis
• Title not available (Why is that?)
• Title not available (Why is that?)
• On the differential geometry of frame bundles of Riemannian manifolds.
• Horizontal dimensionality reduction and iterated frame bundle development
• Simulation of conditioned diffusion and application to parameter estimation
• Title not available (Why is that?)
• Barycentric subspace analysis on manifolds
• A geometric framework for stochastic shape analysis
• Modelling anisotropic covariance using stochastic development and sub-Riemannian frame bundle geometry
• Differential geometry and stochastic dynamics with deep learning numerics
• Evolution Equations with Anisotropic Distributions and Diffusion PCA

Cited In (11)

• Probabilistic approaches to geometric statistics. Stochastic processes, transition distributions, and fiber bundle geometry
• Principal arc analysis on direct product manifolds
• Modelling anisotropic covariance using stochastic development and sub-Riemannian frame bundle geometry
• Recent advances in directional statistics
• Probabilistic geodesic models for regression and dimensionality reduction on Riemannian manifolds
• Statistics for data with geometric structure. Abstracts from the workshop held January 21--27, 2018
• Statistical methods generalizing principal component analysis to non-Euclidean spaces
• Principal flows
• Probabilistic modelling of general noisy multi-manifold data sets
• Local eigenvalue decomposition for embedded Riemannian manifolds
• Scale and curvature effects in principal geodesic analysis

Uses Software

• Theano

This page was built for publication: An infinitesimal probabilistic model for principal component analysis of manifold valued data

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2316992)