Unsupervised Deep Manifold Attributed Graph Embedding

TL;DR

This paper introduces DMAGE, a novel unsupervised graph embedding method that effectively captures structural and feature information, reduces oversmoothing, and improves performance on visualization, clustering, and link prediction tasks.

Contribution

The paper proposes a new framework using geodesic similarity and Bergman divergence, along with a network design to mitigate oversmoothing and enhance stability in attributed graph embedding.

Findings

DMAGE outperforms existing methods on multiple downstream tasks.

The approach effectively reduces oversmoothing in deep graph models.

It achieves significant improvements across four datasets.

Abstract

Unsupervised attributed graph representation learning is challenging since both structural and feature information are required to be represented in the latent space. Existing methods concentrate on learning latent representation via reconstruction tasks, but cannot directly optimize representation and are prone to oversmoothing, thus limiting the applications on downstream tasks. To alleviate these issues, we propose a novel graph embedding framework named Deep Manifold Attributed Graph Embedding (DMAGE). A node-to-node geodesic similarity is proposed to compute the inter-node similarity between the data space and the latent space and then use Bergman divergence as loss function to minimize the difference between them. We then design a new network structure with fewer aggregation to alleviate the oversmoothing problem and incorporate graph structure augmentation to improve the representation's stability. Our proposed DMAGE surpasses state-of-the-art methods by a significant margin on three downstream tasks: unsupervised visualization, node clustering, and link prediction across four popular datasets.