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

Abstract: In this article, we study the whole theory of regularized learning for linear-functional data in Banach spaces including representer theorems, pseudo-approximation theorems, and convergence theorems. The input training data are composed of linear functionals in the predual space of the Banach space to represent the discrete local information of multimodel data and multiscale models. The training data and the multi-loss functions are used to compute the empirical risks to approximate the expected risks, and the regularized learning is to minimize the regularized empirical risks over the Banach spaces. The exact solutions of the original problems are approximated globally by the regularized learning even if the original problems are unknown or unformulated. In the convergence theorems, we show the convergence of the approximate solutions to the exact solutions by the weak* topology of the Banach space. Moreover, the theorems of the regularized learning are applied to solve many problems of machine learning such as support vector machines and artificial neural networks.