Bounds on the Approximation Power of Feedforward Neural Networks
Machine Learning
2018-07-02 v1 Information Theory
math.IT
Machine Learning
Abstract
The approximation power of general feedforward neural networks with piecewise linear activation functions is investigated. First, lower bounds on the size of a network are established in terms of the approximation error and network depth and width. These bounds improve upon state-of-the-art bounds for certain classes of functions, such as strongly convex functions. Second, an upper bound is established on the difference of two neural networks with identical weights but different activation functions.
Keywords
Cite
@article{arxiv.1806.11416,
title = {Bounds on the Approximation Power of Feedforward Neural Networks},
author = {Mohammad Mehrabi and Aslan Tchamkerten and Mansoor I. Yousefi},
journal= {arXiv preprint arXiv:1806.11416},
year = {2018}
}