Analyzing Heterogeneous Networks With Missing Attributes by Unsupervised Contrastive Learning

Heterogeneous information networks (HINs) are potent models of complex systems. In practice, many nodes in an HIN have their attributes unspecified, resulting in significant performance degradation for supervised and unsupervised representation learning. We developed an unsupervised heterogeneous graph contrastive learning approach for analyzing HINs with missing attributes (HGCA). HGCA adopts a contrastive learning strategy to unify attribute completion and representation learning in an unsupervised heterogeneous framework. To deal with a large number of missing attributes and the absence of labels in unsupervised scenarios, we proposed an augmented network to capture the semantic relations between nodes and attributes to achieve a fine-grained attribute completion. Extensive experiments on three large real-world HINs demonstrated the superiority of HGCA over several state-of-the-art methods. The results also showed that the complemented attributes by HGCA can improve the performance of existing HIN models.

I. Introduction

Many real-world systems, e.g., transportation networks, power grids, and social networks, are best viewed and formulated as networks. The overarching problem of mining and analyzing valuable information in networks has been actively studied for decades [1], [2]. As a more capable representation scheme using multiple types of nodes and edges, heterogeneous information networks (HINs) were recently introduced to model complex systems with various types of entities and relations [3]. Low-dimensional embedding techniques have also been adopted to derive compact representations of HINs and extract network-specific information, such as heterogeneous network structural properties, and node semantic relations [3]. Several methods have been developed for heterogeneous network embedding (HNE), including proximity-preserving methods, message-passing methods, and relation-learning methods [3]. Among these methods are the popular ones based on heterogeneous graph neural networks (HGNNs), which have been applied to, e.g., node classification [4], [5] and link prediction [6], [7].