MPXGAT: An Attention based Deep Learning Model for Multiplex Graphs Embedding

From transportation systems to power grids, from the network of our social relationships to that of neurons in our brains, complex networks are all around us. Due to such ubiquity, network and graph theory have imposed themselves in many research fields, from engineering to physics, social science, and biology [1, 2, 3, 4]. A topic that has recently received considerable interest in computer science is that of how to efficiently represent large-scale graphs [5, 6, 7]. Particularly, graph embedding methods, which consist in projecting the elements of a graph, i.e., vertices, edges, and motifs, to a low-dimensional vector space by preserving some of the graph properties, have shown to be very successful in graph representation [8]. These embedding techniques are suitable for multiple applications, as they can be used in downstream learning tasks, including node classification [9], link prediction [10], and community detection [11].