Leveraging Non-linear Dimension Reduction and Random Walk Co-occurrence for Node Embedding

TL;DR

This paper introduces COVE, a high-dimensional, explainable node embedding method that leverages non-linear dimension reduction and random walk co-occurrence, improving clustering and link prediction performance.

Contribution

It proposes a novel high-dimensional embedding approach, COVE, that is compatible with UMAP for dimension reduction and demonstrates competitive community detection results.

Findings

COVE slightly outperforms traditional embeddings on clustering tasks.

The COVE UMAP HDBSCAN pipeline performs comparably to Louvain.

Embedding is inspired by neural methods using co-occurrence and diffusion processes.

Abstract

Leveraging non-linear dimension reduction techniques, we remove the low dimension constraint from node embedding and propose COVE, an explainable high dimensional embedding that, when reduced to low dimension with UMAP, slightly increases performance on clustering and link prediction tasks. The embedding is inspired by neural embedding methods that use co-occurrence on a random walk as an indication of similarity, and is closely related to a diffusion process. Extending on recent community detection benchmarks, we find that a COVE UMAP HDBSCAN pipeline performs similarly to the popular Louvain algorithm.