Interplay between depth of neural networks and locality of target functions

Autor:	Mori, Takashi, Ueda, Masahito
Rok vydání:	2022
Předmět:	Computer Science - Machine Learning Condensed Matter - Disordered Systems and Neural Networks Statistics - Machine Learning
Druh dokumentu:	Working Paper
Popis:	It has been recognized that heavily overparameterized deep neural networks (DNNs) exhibit surprisingly good generalization performance in various machine-learning tasks. Although benefits of depth have been investigated from different perspectives such as the approximation theory and the statistical learning theory, existing theories do not adequately explain the empirical success of overparameterized DNNs. In this work, we report a remarkable interplay between depth and locality of a target function. We introduce $k$-local and $k$-global functions, and find that depth is beneficial for learning local functions but detrimental to learning global functions. This interplay is not properly captured by the neural tangent kernel, which describes an infinitely wide neural network within the lazy learning regime. Comment: 15 pages. This paper is a revised version of arXiv:2005.12488
Databáze:	arXiv
Externí odkaz:	http://arxiv.org/abs/2201.12082 Zobrazit plný text záznamu View this record from Arxiv