רמי פוזיס

Developing a tailor-made centrality measure for a given task requires domain and network analysis expertise, as well as time and effort. Automatically learning arbitrary centrality measures provided ground truth node scores is an important research direction. In this article, we propose a generic deep learning architecture for centrality learning that relies on the insight that arbitrary centrality measures can be computed using Routing Betweenness Centrality (RBC) and our new differentiable implementation of RBC. The proposed Learned Routing Centrality (LRC) architecture optimizes the routing function of RBC to fit the ground truth scores. Results show that LRC can learn multiple types of centrality indices more accurately than state-of-the-art.