Deep Partial Multiplex Network Embedding

Abstract

Network embedding is an effective technique to learn the low-dimensional representations of nodes in networks. Realworld networks are usually with multiplex or having multi-view representations from different relations. Recently, there has been increasing interest in network embedding on multiplex data. However, most existing multiplex approaches assume that the data is complete in all views. But in real applications, it is often the case that each view suffers from the missing of some data and therefore results in partial multiplex data.In this paper, we present a novel Deep Partial Multiplex Network Embedding approach to deal with incomplete data. In particular, the network embeddings are learned by simultaneously minimizing the deep reconstruction loss with the autoencoder neural network, enforcing the data consistency across views via common latent subspace learning, and preserving the data topological structure within the same network through graph Laplacian. We further prove the orthogonal invariant property of the learned embeddings and connect our approach with the binary embedding techniques. Experiments on four multiplex benchmarks demonstrate the superior performance of the proposed approach over several stateof-the-art methods on node classification, link prediction and clustering tasks. CCS CONCEPTS• Information systems → Web searching and information discovery.

CS-MLGCN: Multiplex Graph Convolutional Networks for Community Search in Multiplex Networks

“…One idea to overcome this challenge is to treat each view equally and incorporate the information of all views. However, in many real-world multiplex graphs, the data in some types of relations are noisy/insignificant [18,37], indicating lesser importance.…”

“…(2) These methods treat each relation type (aka view) equally for identifying the community, while many real-world multiplex networks may contain noisy/insignificant views [18]. Moreover, all nodes in the network might not have full information in all views, and so these noisy/insignificant views may be different for each vertex [18,37]. (3) These methods only consider the structural properties of the network, while in realworld applications, networks often come naturally endowed with attributes associated with their vertices.…”