Adaptive Network Alignment with Unsupervised and Multi-order Convolutional Networks

Abstract

Network alignment is the problem of pairing nodes between two graphs such that the paired nodes are structurally and semantically similar. A well-known application of network alignment is to identify which accounts in different social networks belong to the same person. Existing alignment techniques, however, lack scalability, cannot incorporate multi-dimensional information without training data, and are limited in the consistency constraints enforced by an alignment. In this paper, we propose a fully unsupervised network alignment framework based on a multi-order embedding model. The model learns the embeddings of each node using a graph convolutional neural representation, which we prove to satisfy consistency constraints. We further design a data augmentation method and a refinement mechanism to make the model adaptive to consistency violations and noise. Extensive experiments on real and synthetic datasets show that our model outperforms state-of-the-art alignment techniques. We also demonstrate the robustness of our model against adversarial conditions, such as structural noises, attribute noises, graph size imbalance, and hyper-parameter sensitivity.

Network Alignment with Holistic Embeddings

“…DeepLink [29] employs an auto-encoder to better construct the mapping function. GAlign [30] exploits the multi-order nature of a graph convolutional network [17] to simultaneously integrate the attributional and topological consistency constraints to tackle the alignment task. However, these methods rely solely on network topology and fail to exploit the richness of network node information, and thus remain susceptible to structural noise in real-world networks.…”

“…attributional information, high-order proximity, community structure). The first approach, GCNbased embedding, is the improvement from our previous work [30], where we redesign the loss module to be appropriate with the supervised setting and thus facilitate the unifying with other embeddings. The second approach, global community-aware embedding, goes beyond the existing works in graph community detection [31], [32], [33] by encoding the community information in nodes' embedding.…”

“…Many existing embedding-based network aligners fail to satisfy these two consistency constraints, due to differences in the modality of the information type and the limits on the richness of the information incorporated in the embedding [14]. Although a few techniques have attempted to cover both types of constraints, they consider them in separate steps [25] or overlook valuable information [30].…”

“…This removes the requirement for a reconciliation step for the two embedding spaces. In addition, this mechanism ensures that the consistency constraints are satisfied, since corresponding nodes with the same structural and attributional information propagated with shared weights will have identical embeddings [30].…”

Çoklu Sosyal Ağlarda Aynı Kullanıcıları Belirleme Yöntemi

“…Zhang ve arkadaşları topoloji tabanlı hizalama sürecine rehberlik etmek için düğüm nitelik bilgisinden faydalanan bir düğüm hizalama yöntemi önermişlerdir [22]. Trung ve arkadaşları gömme modeline dayanan tamamen denetlenmeyen bir ağ hizalama çerçevesi önermiştir [23]. Zhan ve arkadaşları birleşik bir bağlantı tahmini çerçevesi önererek toplu bağlantı füzyonunu (CLF) temel alan kolektif rastgele yürüyüş ile kısmen hizalanmış olasılıksal ağlar kullanmışlardır [24].…”