Principal Component Analysis for Functional Data on Riemannian Manifolds and Spheres

Xiongtao Dai,Hans-Georg Müller

2017-10-24

Abstract:Functional data analysis on nonlinear manifolds has drawn recent interest. Sphere-valued functional data, which are encountered for example as movement trajectories on the surface of the earth, are an important special case. We consider an intrinsic principal component analysis for smooth Riemannian manifold-valued functional data and study its asymptotic properties. Riemannian functional principal component analysis (RFPCA) is carried out by first mapping the manifold-valued data through Riemannian logarithm maps to tangent spaces around the time-varying Fréchet mean function, and then performing a classical multivariate functional principal component analysis on the linear tangent spaces. Representations of the Riemannian manifold-valued functions and the eigenfunctions on the original manifold are then obtained with exponential maps. The tangent-space approximation through functional principal component analysis is shown to be well-behaved in terms of controlling the residual variation if the Riemannian manifold has nonnegative curvature. Specifically, we derive a central limit theorem for the mean function, as well as root-$n$ uniform convergence rates for other model components, including the covariance function, eigenfunctions, and functional principal component scores. Our applications include a novel framework for the analysis of longitudinal compositional data, achieved by mapping longitudinal compositional data to trajectories on the sphere, illustrated with longitudinal fruit fly behavior patterns. RFPCA is shown to be superior in terms of trajectory recovery in comparison to an unrestricted functional principal component analysis in applications and simulations and is also found to produce principal component scores that are better predictors for classification compared to traditional functional functional principal component scores.

Statistics Theory

What problem does this paper attempt to address?

The paper is primarily dedicated to addressing the problem of functional data analysis on Riemannian manifolds, particularly how to perform Principal Component Analysis (PCA) on datasets with nonlinear geometric structures. Specifically, the goals of the paper include:

1. **Proposing an intrinsic Riemannian Functional Principal Component Analysis (RFPCA) method**: By mapping data on Riemannian manifolds to the tangent space and performing traditional multivariate Functional Principal Component Analysis (FPCA) in these tangent spaces, the method achieves dimensionality reduction while preserving the inherent geometric properties of the data.

2. **Demonstrating the asymptotic properties of the method**: The authors derive the convergence rates between the sample estimators and the population parameters for the mean function, covariance function, eigenfunctions, and functional principal component scores. They also show that when the Riemannian manifold has non-negative curvature, the FPCA based on the tangent space can effectively control residual variation.

3. **Application to specific types of data**: In particular, the paper discusses in detail the case when the data lie on the Euclidean sphere, i.e., Spherical Functional Principal Component Analysis (SFPCA), and demonstrates how this method can be applied to the analysis of longitudinal compositional data.

4. **Theoretical and practical validation**: Through simulation studies and applications to real datasets (such as airplane flight trajectories and fruit fly behavioral patterns), it is shown that RFPCA outperforms unconstrained FPCA in trajectory recovery and that the resulting principal component scores perform better in classification tasks.

In summary, the main objective of this paper is to develop an efficient dimensionality reduction technique for functional data on manifolds and to demonstrate its advantages in various application scenarios.