dyngraph2vec: Capturing Network Dynamics using Dynamic Graph Representation Learning

By Palash Goyal, Sujit Rokka Chhetri, Arquimedes Canedo. Knowledge-Based Systems, 2018

454 citations recorded by the source corpus.

Abstract Learning graph representations is a fundamental task aimed at capturing various properties of graphs in vector space. The most recent methods learn such representations for static networks. However, real-world networks evolve over time and have varying dynamics. Capturing such evolution is key to predicting the properties of unseen networks. To understand how the network dynamics affect the prediction performance, we propose an embedding approach which learns the structure of evolution in dynamic graphs and can predict unseen links with higher precision. Our model, dyngraph2vec , learns the temporal transitions in the network using a deep architecture composed of dense and recurrent layers. We motivate the need for capturing dynamics for the prediction on a toy data set created using stochastic block models. We then demonstrate the efficacy of dyngraph2vec over existing state-of-the-art methods on two real-world data sets. We observe that learning dynamics can improve the quality of embedding and yield better performance in link prediction.

Subfields: Artificial Intelligence, Experimental and Cognitive Psychology, Statistical and Nonlinear Physics

View Paper: https://www.semanticscholar.org/paper/f6e59062382fdec9b95c3abef1c27efc3b2ec1c7

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

In Experimental and Cognitive Psychology

In Statistical and Nonlinear Physics

Citations in the corpus

Foundations and Modeling of Dynamic Networks Using Dynamic Graph Neural Networks: A Survey (2020, methodology): “This approach is used by several models for link prediction [16], [47], [101].”
Recurrent Event Network: Autoregressive Structure Inference over Temporal Knowledge Graphs (2019, methodology): “There are attempts on predicting future links on homogeneous graphs (Pareja et al., 2020; Goyal et al., 2018, 2019; Zhou et al., 2018; Singer et al., 2019).”
Relational Representation Learning for Dynamic (Knowledge) Graphs: A Survey (2019, background): “To take more snapshots into account in learning node embeddings, Goyal et al. [82] propose to learn a single autoencoder where at snapshot t , the ﬁrst component of the autoencoder takes as input A t − l [ i ] , A t − l +1 [ i ] , . . . , A t − 1 [ i ] and produces a vector z t v i corresponding to…”
Inductive Representation Learning in Temporal Networks via Causal Anonymous Walks (2021, methodology): “, 2017) to encode structural patterns and temporal patterns respectively (Pareja et al., 2020; Manessi et al., 2020; Goyal et al., 2020; Hajiramezanali et al., 2019; Sankar et al., 2020).”
DySAT: Deep Neural Representation Learning on Dynamic Graphs via Self-Attention Networks (2020, methodology): “Existing dynamic graph representation learning methods mainly fall into categories: temporal regularizers that enforce smoothness of node representations from adjacent snapshots [39, 40], and recurrent neural networks [6, 9] that summarize historical snapshots via hidden states.”