Classifying Network Data with Deep Kernel Machines

TL;DR

This paper introduces deep kernel machines, a nonlinear extension of traditional kernel classifiers, demonstrating improved node classification performance on graphs by stacking nonlinear transformations in the feature space.

Contribution

It proposes the concept of deep kernel machines, extending kernel methods with multiple nonlinear layers, and shows their effectiveness for network node classification.

Findings

Deep kernel machines outperform traditional kernel classifiers in graph node classification.

Nonlinear transformations in the feature space significantly improve classification accuracy.

Connections are drawn between deep kernel machines and recent deep learning architectures.

Abstract

Inspired by a growing interest in analyzing network data, we study the problem of node classification on graphs, focusing on approaches based on kernel machines. Conventionally, kernel machines are linear classifiers in the implicit feature space. We argue that linear classification in the feature space of kernels commonly used for graphs is often not enough to produce good results. When this is the case, one naturally considers nonlinear classifiers in the feature space. We show that repeating this process produces something we call "deep kernel machines." We provide some examples where deep kernel machines can make a big difference in classification performance, and point out some connections to various recent literature on deep architectures in artificial intelligence and machine learning.