A comparison of deep networks with ReLU activation function and linear spline-type methods

TL;DR

This paper compares the expressive power of deep neural networks with ReLU activation to spline methods, showing neural networks can efficiently approximate spline-based functions and perform comparably in statistical risk.

Contribution

It demonstrates that DNNs can properly learn MARS and Faber-Schauder spline functions with efficient parameter complexity, providing theoretical bounds and constructive proofs.

Findings

DNNs can approximate MARS functions with O(M log(M/ε)) parameters.

DNNs can approximate Faber-Schauder expansions similarly.

Neural networks perform as well or slightly better than spline methods in statistical risk.

Abstract

Deep neural networks (DNNs) generate much richer function spaces than shallow networks. Since the function spaces induced by shallow networks have several approximation theoretic drawbacks, this explains, however, not necessarily the success of deep networks. In this article we take another route by comparing the expressive power of DNNs with ReLU activation function to piecewise linear spline methods. We show that MARS (multivariate adaptive regression splines) is improper learnable by DNNs in the sense that for any given function that can be expressed as a function in MARS with $M$ parameters there exists a multilayer neural network with $O(M \log (M/\varepsilon))$ parameters that approximates this function up to sup-norm error $\varepsilon.$ We show a similar result for expansions with respect to the Faber-Schauder system. Based on this, we derive risk comparison inequalities that bound the statistical risk of fitting a neural network by the statistical risk of spline-based methods. This shows that deep networks perform better or only slightly worse than the considered spline methods. We provide a constructive proof for the function approximations.