Effective Visualization of Information Diffusion Process over Complex Networks

Effective visualization is vital for understanding a complex network, in particular its dynamical aspect such as information diffusion process. Existing node embedding methods are all based solely on the network topology and sometimes produce counter-intuitive visualization. A new node embedding method based on conditional probability is proposed that explicitly addresses diffusion process using either the IC or LT models as a cross-entropy minimization problem, together with two label assignment strategies that can be simultaneously adopted. Numerical experiments were performed on two large real networks, one represented by a directed graph and the other by an undirected graph. The results clearly demonstrate the advantage of the proposed methods over conventional spring model and topology-based cross-entropy methods, especially for the case of directed networks.