A comparison of deep networks with ReLU activation function and linear spline-type methods
This provides theoretical insights into why deep networks succeed, addressing a foundational problem in machine learning theory.
The paper tackles the problem of understanding the expressive power of deep neural networks (DNNs) with ReLU activation by comparing them to piecewise linear spline methods, showing that DNNs can approximate functions from MARS with O(M log(M/ε)) parameters up to sup-norm error ε and perform similarly or better in statistical risk.
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.