Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Representation Learning over Dynamic Graphs

About

How can we effectively encode evolving information over dynamic graphs into low-dimensional representations? In this paper, we propose DyRep, an inductive deep representation learning framework that learns a set of functions to efficiently produce low-dimensional node embeddings that evolves over time. The learned embeddings drive the dynamics of two key processes namely, communication and association between nodes in dynamic graphs. These processes exhibit complex nonlinear dynamics that evolve at different time scales and subsequently contribute to the update of node embeddings. We employ a time-scale dependent multivariate point process model to capture these dynamics. We devise an efficient unsupervised learning procedure and demonstrate that our approach significantly outperforms representative baselines on two real-world datasets for the problem of dynamic link prediction and event time prediction.

Rakshit Trivedi, Mehrdad Farajtabar, Prasenjeet Biswal, Hongyuan Zha• 2018

Related benchmarks

TaskDatasetResultRank
Inductive dynamic link predictionReddit (inductive)
AUC-ROC (%)95.89
159
Dynamic Link PredictionLastFM (transductive)
AP71.85
143
Dynamic Link PredictionWikipedia (inductive)
AP92.21
119
Inductive dynamic link predictionWikipedia (inductive)
AUC-ROC0.9221
116
transductive dynamic link predictionWikipedia
AUC ROC94.43
109
transductive dynamic link predictionREDDIT
AUC-ROC0.9813
105
transductive dynamic link predictionSocial Evo.
AUC ROC90.37
105
Dynamic Link PredictionReddit (transductive)--
92
Dynamic Link PredictionMOOC (transductive)--
80
Inductive dynamic link predictionLastFM
AUC-ROC83.47
73
Showing 10 of 55 rows

Other info

Follow for update