Social network analysis: incorporating real world message passing in GRLs

Abstract

With social media becoming the primary source of how information is created, consumed and distributed, it is essential for the learning models to utilize the flow in order to get better understanding which can be later used for downstream tasks. Graphs Are the most effective structure which can model real life information cascades in social networks. But traditional machine learning algorithms are ineffective in modeling knowledge flow in large dynamic networks. We present to you modified versions of existing GRL algorithms which can utilize the flow of information efficiently and will give us the embedding of the nodes updated dynamically with time. These embeddings then can be used in our downstream tasks.