Constructing traffic-aware overlay topologies: a machine learning approach

Recent game-theoretic approaches to constructing overlay network topologies have not been scalable. This paper introduces a machine learning approach to constructing overlay networks. The machine learning approach learns characteristics from small networks constructed using a game-theoretic approach. The knowledge learned is then used to construct larger networks. The results show that the machine learning approach closely approximates the game-theoretic networks for a wide range of network parameters, while being scalable.