Large-width functional asymptotics for deep Gaussian neural networks

MaRDI QIDQ6361137zbMATH OpenFDO

Authors Daniele Bracale, Stefano Favaro, S. Fortini, Stefano Peluchetti

Publication date 20 February 2021

Abstract: In this paper, we consider fully connected feed-forward deep neural networks where weights and biases are independent and identically distributed according to Gaussian distributions. Extending previous results (Matthews et al., 2018a;b; Yang, 2019) we adopt a function-space perspective, i.e. we look at neural networks as infinite-dimensional random elements on the input space $mathbb{R}^I$. Under suitable assumptions on the activation function we show that: i) a network defines a continuous Gaussian process on the input space $mathbb{R}^I$; ii) a network with re-scaled weights converges weakly to a continuous Gaussian process in the large-width limit; iii) the limiting Gaussian process has almost surely locally $gamma$-H"older continuous paths, for $0<gamma<1$. Our results contribute to recent theoretical studies on the interplay between infinitely wide deep neural networks and Gaussian processes by establishing weak convergence in function-space with respect to a stronger metric.

This page was built for publication: Large-width functional asymptotics for deep Gaussian neural networks

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6361137)