Tag2Gauss: Learning Tag Representations via Gaussian Distribution in Tagged Networks

Abstract

Keyword-based tags (referred to as tags) are used to represent additional attributes of nodes in addition to what is explicitly stated in their contents, like the hashtags in YouTube. Aside of being auxiliary information for node representation, tags can also be used for retrieval, recommendation, content organization, and event analysis. Therefore, tag representation learning is of great importance. However, to learn satisfactory tag representations is challenging because 1) traditional representation methods generally fail when it comes to representing tags, 2) bidirectional interactions between nodes and tags should be modeled, which are generally not dealt within existing research works. In this paper, we propose a tag representation learning model which takes tag-related node interaction into consideration, named Tag2Gauss. Specifically, since tags represent node communities with intricate overlapping relationships, we propose that Gaussian distributions would be appropriate in modeling tags. Considering the bidirectional interactions between nodes and tags, we propose a tag representation learning model mapping tags to distributions consisting of two embedding tasks, namely Tag-view embedding and Node-view embedding. Extensive evidence demonstrates the effectiveness of representing tag as a distribution, and the advantages of the proposed architecture in many applications, such as the node classification and the network visualization.

Understanding and Improvement of Adversarial Training for Network Embedding from an Optimization Perspective

“…On the other hand, adversarial perturbations on model parameters demonstrates an entirely new paradigm of adversarial training and shows great potential in promoting model effectiveness, especially for those solving transductive embedding problems. A reasonable explanation is in urgent need to exhibit why and where APP can work well, and thus we are able to determine whether APP can be generalized to other similar scenarios such as word embedding [25,30], tag embedding [45,47], table embedding [13,18,52] and broader domains.…”

HTGN-BTW: Heterogeneous Temporal Graph Network with Bi-Time-Window Training Strategy for Temporal Link Prediction

“…Graph, as a general data structure to represent relationships between entities, is ubiquitous in the real world. Many of them are temporal graphs that are dynamic and evolving over time [3,21], such as E-commerce networks, social networks and communication networks [13,15]. In such real graphs, one popular demand is temporal link prediction which is to predict new links in the future according to the historical information, such as recommending products to users in E-commerce networks, recommending friends to users in social networks, etc [2,11].…”

Fast Hierarchy Preserving Graph Embedding via Subspace Constraints

“…† Corresponding Authors. 1 CIKM19 Best Research Paper Runner-up Award found in the social network of animals, and category hierarchy is widely used in lots of e-commerce sites (e.g., Alibaba, Amazon and Rakuten Ichiba) [11,12,13]. "Women's Fashion → Tops → Sweaters → Longsleeved knit → Crew neck" is an example of such an organization.…”