Abstract:We consider the problem of estimating the slope function in a functional regression with a scalar response and a functional covariate. This central problem of functional data analysis is well known to be ill-posed, thus requiring a regularised estimation procedure. The two most commonly used approaches are based on spectral truncation or Tikhonov regularisation of the empirical covariance operator. In principle, Tikhonov regularisation is the more canonical choice. Compared to spectral truncation, it is robust to eigenvalue ties, while it attains the optimal minimax rate of convergence in the mean squared sense, and not just in a concentration probability sense. In this paper, we show that, surprisingly, one can strictly improve upon the performance of the Tikhonov estimator in finite samples by means of a linear estimator, while retaining its stability and asymptotic properties by combining it with a form of spectral truncation. Specifically, we construct an estimator that additively decomposes the functional covariate by projecting it onto two orthogonal subspaces defined via functional PCA; it then applies Tikhonov regularisation to the one component, while leaving the other component unregularised. We prove that when the covariate is Gaussian, this hybrid estimator uniformly improves upon the MSE of the Tikhonov estimator in a non-asymptotic sense, effectively rendering it inadmissible. This domination is shown to also persist under discrete observation of the covariate function. The hybrid estimator is linear, straightforward to construct in practice, and with no computational overhead relative to the standard regularisation methods. By means of simulation, it is shown to furnish sizeable gains even for modest sample sizes.

Functional Regularized Least Squares Classi cation with Operator-valued Kernels

Multiple Operator-valued Kernel Learning

Kernel partial least squares regression for functional nonlinear models

Duality in RKHSs with Infinite Dimensional Outputs: Application to Robust Losses

Random Fourier Features for Operator-Valued Kernels

Functional Linear Regression with Mixed Predictors

Nonparametric Operator-Regularized Covariance Function Estimation for Functional Data

Robust Regularized Kernel Regression.

Nonparametric learning of kernels in nonlocal operators

Minimax Optimal Kernel Operator Learning via Multilevel Training

Analysis of Regularized Learning for Linear-functional Data in Banach Spaces

Sparse multiresolution representations with adaptive kernels

Statistical Optimality of Divide and Conquer Kernel-based Functional Linear Regression

Low-rank kernel regression with preserved locality for multi-class analysis

An RKHS model for variable selection in functional regression

M-Power Regularized Least Squares Regression

Online Regularized Learning Algorithm for Functional Data

Hybrid Regularisation of Functional Linear Models

Learning Analysis of Kernel Ridgeless Regression with Asymmetric Kernel Learning

Classification of Functional Data with k-Nearest-Neighbor Ensembles by Fitting Constrained Multinomial Logit Models

A Generalized Kernel Approach to Structured Output Learning