Finite Neural Networks as Mixtures of Gaussian Processes: From Provable Error Bounds to Prior Selection

TL;DR

This paper introduces a method to approximate finite neural networks with mixtures of Gaussian processes, providing error bounds and enabling better understanding and tuning of neural network behavior.

Contribution

The authors develop an algorithmic framework to approximate finite neural networks with mixtures of Gaussian processes, including error bounds and applications to prior selection.

Findings

The method provides an $oldsymbol{ extepsilon}$-close approximation at finite input points.

It enables tuning neural networks to mimic Gaussian process behavior.

Empirical results demonstrate effectiveness on regression and classification tasks.

Abstract

Infinitely wide or deep neural networks (NNs) with independent and identically distributed (i.i.d.) parameters have been shown to be equivalent to Gaussian processes. Because of the favorable properties of Gaussian processes, this equivalence is commonly employed to analyze neural networks and has led to various breakthroughs over the years. However, neural networks and Gaussian processes are equivalent only in the limit; in the finite case there are currently no methods available to approximate a trained neural network with a Gaussian model with bounds on the approximation error. In this work, we present an algorithmic framework to approximate a neural network of finite width and depth, and with not necessarily i.i.d. parameters, with a mixture of Gaussian processes with error bounds on the approximation error. In particular, we consider the Wasserstein distance to quantify the closeness between probabilistic models and, by relying on tools from optimal transport and Gaussian processes, we iteratively approximate the output distribution of each layer of the neural network as a mixture of Gaussian processes. Crucially, for any NN and $\epsilon >0$ our approach is able to return a mixture of Gaussian processes that is $\epsilon$-close to the NN at a finite set of input points. Furthermore, we rely on the differentiability of the resulting error bound to show how our approach can be employed to tune the parameters of a NN to mimic the functional behavior of a given Gaussian process, e.g., for prior selection in the context of Bayesian inference. We empirically investigate the effectiveness of our results on both regression and classification problems with various neural network architectures. Our experiments highlight how our results can represent an important step towards understanding neural network predictions and formally quantifying their uncertainty.