Deep learning with differential Gaussian process flows

TL;DR

This paper introduces differential Gaussian process flows, a new deep learning approach that models input transformations as stochastic differential equations, achieving superior performance over existing methods.

Contribution

It presents a novel framework combining stochastic differential equations with Gaussian processes for deep learning, extending discrete layers into a continuous dynamical system.

Findings

Achieves state-of-the-art results surpassing deep Gaussian processes and neural networks.

Demonstrates the effectiveness of differential Gaussian process flows in classification and regression tasks.

Introduces a new paradigm for deep learning based on stochastic differential equations.

Abstract

We propose a novel deep learning paradigm of differential flows that learn a stochastic differential equation transformations of inputs prior to a standard classification or regression function. The key property of differential Gaussian processes is the warping of inputs through infinitely deep, but infinitesimal, differential fields, that generalise discrete layers into a dynamical system. We demonstrate state-of-the-art results that exceed the performance of deep Gaussian processes and neural networks