# Variational Graph Recurrent Neural Networks Source: [Variational Graph Recurrent Neural Networks]() Collection index: [llms.txt]() Data updated: 2026-01-04T03:07:12.455Z Authors: Ehsan Hajiramezanali, Arman Hasanzadeh, Nick Duffield, Krishna R Narayanan, Mingyuan Zhou, Xiaoning Qian Published: 2019-08-26T14:44:47.000Z ## Abstract Representation learning over graph structured data has been mostly studied in static graph settings while efforts for modeling dynamic graphs are still scant. In this paper, we develop a novel hierarchical variational model that introduces additional latent random variables to jointly model the hidden states of a graph recurrent neural network (GRNN) to capture both topology and node attribute changes in dynamic graphs. We argue that the use of high-level latent random variables in this variational GRNN (VGRNN) can better capture potential variability observed in dynamic graphs as well as the uncertainty of node latent representation. With semi-implicit variational inference developed for this new VGRNN architecture (SI-VGRNN), we show that flexible non-Gaussian latent representations can further help dynamic graph analytic tasks. Our experiments with multiple real-world dynamic graph datasets demonstrate that SI-VGRNN and VGRNN consistently outperform the existing baseline and state-of-the-art methods by a significant margin in dynamic link prediction. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Dynamic Link Detection on Enron](): Dynamic Link Detection; ENRON; AP: 96.31. - [Dynamic New Link Prediction on COLAB](): Dynamic new link prediction; COLAB; AUC: 0.8595. - [Dynamic new link prediction on Facebook](): Dynamic new link prediction; Facebook; AUC: 90.94. - [Dynamic Link Detection on COLAB (10 runs on random splits)](): Dynamic Link Detection; COLAB (10 runs on random splits); AUC: 89.15. - [Dynamic Link Detection on Facebook 10 runs on random splits](): Dynamic Link Detection; Facebook 10 runs on random splits; AUC: 0.8812. - [Dynamic Link Detection on Social Evo. (10 runs on random splits)](): Dynamic Link Detection; Social Evo. (10 runs on random splits); AUC: 83.36. - [Dynamic Link Detection on HEP-TH last 10 snapshots (test)](): Dynamic Link Detection; HEP-TH last 10 snapshots (test); AUC: 91.12. - [Dynamic Link Detection on Cora Includes node attributes (10 runs on random splits)](): Dynamic Link Detection; Cora Includes node attributes (10 runs on random splits); AUC: 94.07. - [Link Prediction on Enron (last three snapshots)](): Link Prediction; Enron (last three snapshots); AUC: 92.68. - [Link Prediction on COLAB (last three snapshots)](): Link Prediction; COLAB (last three snapshots); AUC: 0.8656. - [Link Prediction on Facebook (last three snapshots)](): Link Prediction; Facebook (last three snapshots); AUC: 89.44. - [Dynamic new link prediction on Social Evo.](): Dynamic new link prediction; Social Evo.; AP: 0.774. ## Original sources and implementation [Original paper]() [PDF]() [Implementation]()