Network Embedding with Completely-imbalanced Labels

TL;DR

This paper introduces two novel semi-supervised network embedding methods, RSDNE and RECT, designed to effectively handle networks with completely-imbalanced labels, outperforming existing approaches.

Contribution

The paper proposes RSDNE and RECT, new methods that address the challenge of completely-imbalanced labels in network embedding, incorporating intra/inter-class relations and class-semantic knowledge.

Findings

RSDNE guarantees intra-class similarity and inter-class dissimilarity.

RECT leverages class-semantic knowledge for multi-label, feature-rich networks.

Experimental results show superior performance of both methods on real datasets.

Abstract

Network embedding, aiming to project a network into a low-dimensional space, is increasingly becoming a focus of network research. Semi-supervised network embedding takes advantage of labeled data, and has shown promising performance. However, existing semi-supervised methods would get unappealing results in the completely-imbalanced label setting where some classes have no labeled nodes at all. To alleviate this, we propose two novel semi-supervised network embedding methods. The first one is a shallow method named RSDNE. Specifically, to benefit from the completely-imbalanced labels, RSDNE guarantees both intra-class similarity and inter-class dissimilarity in an approximate way. The other method is RECT which is a new class of graph neural networks. Different from RSDNE, to benefit from the completely-imbalanced labels, RECT explores the class-semantic knowledge. This enables RECT to handle networks with node features and multi-label setting. Experimental results on several real-world datasets demonstrate the superiority of the proposed methods. Code is available at https://github.com/zhengwang100/RECT.