Nonparametric operator-regularized covariance function estimation for functional data
Academic Article
Overview
Research
Identity
Additional Document Info
Other
View All
Overview
abstract
2018 Elsevier B.V. In functional data analysis (FDA), the covariance function is fundamental not only as a critical quantity for understanding elementary aspects of functional data but also as an indispensable ingredient for many advanced FDA methods. A new class of nonparametric covariance function estimators in terms of various spectral regularizations of an operator associated with a reproducing kernel Hilbert space is developed. Despite their nonparametric nature, the covariance estimators are automatically positive semi-definite, which is an essential property of covariance functions, via a one-step procedure. An unconventional representer theorem is established to provide a finite dimensional representation for this class of covariance estimators based on data, although the solutions are searched over infinite dimensional functional spaces. To further achieve a low-rank representation, another desirable property, e.g., for dimension reduction and easy interpretation, the trace-norm regularization is particularly studied, under which an efficient algorithm is developed based on the accelerated proximal gradient method. The outstanding practical performance of the trace-norm-regularized covariance estimator is demonstrated by a simulation study and the analysis of a traffic dataset. Under both fixed and random designs, an excellent rate of convergence is established for a broad class of operator-regularized covariance function estimators, which generalizes both the trace-norm-regularized covariance estimator and other popular alternatives.
