Kernel Node Embeddings

TL;DR

This paper introduces a novel kernel-based matrix factorization method for node embeddings that combines random walk information with kernel functions, improving performance on real-world network tasks.

Contribution

It proposes a weighted matrix factorization model that integrates random walk data with kernel functions for enhanced node representation learning.

Findings

Outperforms baseline node embedding algorithms in downstream tasks

Effective integration of random walk data with kernel functions

Demonstrates superior performance on real-world networks

Abstract

Learning representations of nodes in a low dimensional space is a crucial task with many interesting applications in network analysis, including link prediction and node classification. Two popular approaches for this problem include matrix factorization and random walk-based models. In this paper, we aim to bring together the best of both worlds, towards learning latent node representations. In particular, we propose a weighted matrix factorization model which encodes random walk-based information about the nodes of the graph. The main benefit of this formulation is that it allows to utilize kernel functions on the computation of the embeddings. We perform an empirical evaluation on real-world networks, showing that the proposed model outperforms baseline node embedding algorithms in two downstream machine learning tasks.