Local Fr'echet Regression via RKHS embedding and Its Applications to Data Analysis on Manifolds

Abstract: Local Fr'echet Regression (LFR) is a nonparametric regression method for settings in which the explanatory variable lies in a Euclidean space and the response variable lies in a metric space. It is used to estimate smooth trajectories in general metric spaces from noisy observations of random objects taking values in such spaces. Since metric spaces form a broad class of spaces that often lack algebraic structures such as addition or scalar multiplication characteristics typical of vector spaces the asymptotic theory for conventional random variables cannot be directly applied. As a result, deriving the asymptotic distribution of the LFR estimator is challenging. In this paper, we first extend nonparametric regression models for real-valued responses to Hilbert spaces and derive the asymptotic distribution of the LFR estimator in a Hilbert space setting. Furthermore, we propose a new estimator based on the LFR estimator in a reproducing kernel Hilbert space (RKHS), by mapping data from a general metric space into an RKHS. Finally, we consider applications of the proposed method to data lying on manifolds and construct confidence regions in metric spaces based on the derived asymptotic distribution.