Attributed Network Embedding for Learning in a Dynamic Environment

By Jundong Li, Harsh Dani, Xia Hu, Jiliang Tang, Yi Chang, Huan Liu. International Conference on Information and Knowledge Management, 2017

394 citations recorded by the source corpus.

Network embedding leverages the node proximity manifested to learn a low-dimensional node vector representation for each node in the network. The learned embeddings could advance various learning tasks such as node classification, network clustering, and link prediction. Most, if not all, of the existing works, are overwhelmingly performed in the context of plain and static networks. Nonetheless, in reality, network structure often evolves over time with addition/deletion of links and nodes. Also, a vast majority of real-world networks are associated with a rich set of node attributes, and their attribute values are also naturally changing, with the emerging of new content patterns and the fading of old content patterns. These changing characteristics motivate us to seek an effective embedding representation to capture network and attribute evolving patterns, which is of fundamental importance for learning in a dynamic environment. To our best knowledge, we are the first to tackle this problem with the following two challenges: (1) the inherently correlated network and node attributes could be noisy and incomplete, it necessitates a robust consensus representation to capture their individual properties and correlations; (2) the embedding learning needs to be performed in an online fashion to adapt to the changes accordingly. In this paper, we tackle this problem by proposing a novel dynamic attributed network embedding framework - DANE. In particular, DANE first provides an offline method for a consensus embedding and then leverages matrix perturbation theory to maintain the freshness of the end embedding results in an online manner. We perform extensive experiments on both synthetic and real attributed networks to corroborate the effectiveness and efficiency of the proposed framework.

Subfields: Artificial Intelligence, Statistical and Nonlinear Physics

View Paper: https://www.semanticscholar.org/paper/736e8deabcae7e2f9eb6c41a1bfae1b5270a8dbd

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Shirui Pan
2. Chengqi Zhang
3. Guodong Long
4. Jing Jiang
5. Xia Hu
6. Huan Liu
7. Jure Leskovec
8. Philip S. Yu

In Statistical and Nonlinear Physics

1. Shirui Pan
2. Chengqi Zhang
3. Huan Liu
4. Guodong Long
5. Xingquan Zhu
6. Jie Yin
7. Jing Jiang
8. Xia Hu

Citations in the corpus

• EvolveGCN: Evolving Graph Convolutional Networks for Dynamic Graphs (2019, background): “Hence, DANE (Li et al. 2017) updates the eigenvectors effi-ciently based on the prior ones, rather than computing them from scratch for each new graph.”
• Heterogeneous Graph Neural Network (2019, background): “Further, attributed graph embedding models [14, 15, 34] have been proposed to leverages both graph structure and node attributes for learning node embeddings.”
• Deep Anomaly Detection on Attributed Networks (2019, background): “Apart from traditional plain networks in which only node-to-node interactions are observed, attributed networks also encode a rich set of features for each node [2, 13, 18].”
• Relational Representation Learning for Dynamic (Knowledge) Graphs: A Survey (2019, methodology): “Incremental eigenvalue decomposition [36, 141, 228] is based on perturbation theory.”
• DySAT: Deep Neural Representation Learning on Dynamic Graphs via Self-Attention Networks (2020, background): “A related problem is representation learning in streaming graphs where the objective is to improve efficiency over repeatedly retraining static models, by proposing incremental temporal updates [4, 17, 18, 38].”
• Representation Learning for Dynamic Graphs: A Survey (2019, methodology): “The graph used by Liu et al. (2019) has 5,196 nodes, 171,743 edges, and 6 types of node classes.”
• A Survey of Learning Causality with Data: Problems and Methods (2018, background): “Some special cases of this type of data can be attributed networks [80, 145], time series [38] and marked temporal point process [53].”
• Streaming Graph Neural Networks via Continual Learning (2020, methodology): “Our model belongs to the latter type, with typical examples including DANE [11] and DNE [5].”
• Tracing Fake-News Footprints: Characterizing Social Media Messages by How They Propagate (2018, background): “Recent studies also study and utilize network dynamics by observing the change of social networks over time [19, 21].”
• AliGraph: A Comprehensive Graph Neural Network Platform (2019, background): “Considering both the network structure and node attributes, [32] focuses on updating the top eigenvectors and eigenvalues for the streaming network.”
• Learning Graph Embedding With Adversarial Training Methods (2019, background): “DANE [42] deals with the dynamic environment with an incremental matrix factorization approach, and LANE [43] incorporates”
• Network Representation Learning: A Survey (2017, background): “, [106], [107], [108]) assume that the node”
• Predicting Dynamic Embedding Trajectory in Temporal Interaction Networks (2019, methodology): “Other recent algorithms such as DDNE [29], DANE [28], DynGem [15], Zhu et al.”
• A Survey of Learning Causality with Data (2018, background): “ference between units such as links or temporal inter-dependencies between different data units (samples), represented by a matrixA. Some special cases of this type of data can be attributed networks [80, 145], time series [38] and marked temporal point process [53]. Moreover, when there are unobser”
• Deep Representation Learning for Social Network Analysis (2019, methodology): “In addition, DANE (Li et al., 2017b) leverages matrix perturbation theory for tackling online embedding updates.”