Abstract
Low-dimensional vector representations of network nodes have provensuccessful to feed graph data to machine learning algorithms and to improveperformance across diverse tasks. Most of the embedding techniques, however,have been developed with the goal of achieving dense, low-dimensional encodingof network structure and patterns. Here, we present a node embedding techniqueaimed at providing low-dimensional feature vectors that are informative ofdynamical processes occurring over temporal networks - rather than of thenetwork structure itself - with the goal of enabling prediction tasks relatedto the evolution and outcome of these processes. We achieve this by using amodified supra-adjacency representation of temporal networks and building onstandard embedding techniques for static graphs based on random-walks. We showthat the resulting embedding vectors are useful for prediction tasks related toparadigmatic dynamical processes, namely epidemic spreading over empiricaltemporal networks. In particular, we illustrate the performance of our approachfor the prediction of nodes' epidemic states in a single instance of thespreading process. We show how framing this task as a supervised multi-labelclassification task on the embedding vectors allows us to estimate the temporalevolution of the entire system from a partial sampling of nodes at randomtimes, with potential impact for nowcasting infectious disease dynamics.