Statistical Learnability of Generalized Additive Models based on Total Variation Regularization

Matsushima, Shin

arXiv.org Machine Learning 

A generalized additive model (GAM, Hastie and Tibshirani (1987)) is a nonparametric model by the sum of univariate functions with respect to each explanatory variable, i.e., $f({\mathbf x}) = \sum f_j(x_j)$, where $x_j\in\mathbb{R}$ is $j$-th component of a sample ${\mathbf x}\in \mathbb{R}^p$. In this paper, we introduce the total variation (TV) of a function as a measure of the complexity of functions in $L^1_{\rm c}(\mathbb{R})$-space. Our analysis shows that a GAM based on TV-regularization exhibits a Rademacher complexity of $O(\sqrt{\frac{\log p}{m}})$, which is tight in terms of both $m$ and $p$ in the agnostic case of the classification problem. In result, we obtain generalization error bounds for finite samples according to work by Bartlett and Mandelson (2002).

Duplicate Docs Excel Report

Title
None found

Similar Docs  Excel Report  more

TitleSimilaritySource
None found