Big Graph Live

  Graph Classification with Minimum DFS Code:  Improving Graph Neural Network Expressivity A summary of the IEEE BigData  2021 research  paper   by  Jhalak Gupta and Arijit Khan  [presented in Machine Learning on Big Data (MLBD 2021), special session of IEEE BigData 2021]. Background: Graph Classification Given a set of graphs with different structures and sizes, the graph classification problem predicts the class labels of unseen graphs [1, 2, 3]. Developing machine learning tools for classifying graphs can be found in cheminformatics [1, 4] and bioinformatics [6], malware detection [7], telecommunication networks, internet-of-things [8], trajectories and social networks [9]. This is challenging because network data contain graphs with different numbers of nodes and edges, and a generic node order is often not available. Graphs do not have regular grid structures, since the neighborhood size of each node differs. The lack of ordered vector represent...

Multi-relation Graph Summarization A summary of the  ACM Transactions on Knowledge Discovery from Data (TKDD) Journal  2021 research  paper   by  Xiangyu Ke, Arijit Khan, and Francesco Bonchi Background: Multi-relation Graphs Multi-relation networks (also known as multi-layer, multiplex, or multi-dimensional networks) are graphs where multiple edges of different types may exist between any pair of nodes [7]. Multi-relation graphs are an expressive model of real-world activities, in which a relation can be a topic in social networks, an interaction type in genetic networks, or a snapshot in temporal graphs. For instance, BioGRID (thebiogrid.org) describes seven different types of genetic interactions between genes in Homo Sapiens. STRING (string-db.org) models protein-to-protein interactions with six types of correlations statistically learned from existing protein databases, revealing that most protein interactions are associated with at least two types of cor...