Graph Representation Learning for Wireless Communications

Wireless networks are inherently graph-structured, in which, graph representation learning can be utilized to solve complex network optimization problems. In learning, feature vectors for each entity the calculated such that they could capture spatial and temporal dependencies their local global neighborhoods. Specifically, neural (GNNs) powerful tools these problems because of expressive reasoning power. this paper, potential GNNs wireless is presented. An overview provided which covers fundamentals concepts as design over graphs, GNNs, principles. The presented via a few exemplary use cases some initial results on GNNbased access point selection cell-free massive Multiple-Input Multiple-Output (MIMO) systems.