Prof. Janusz Holyst presented today a talk on "Finding the source of local information cascades in complex networks," co-authored by Krzysztof Suchecki, Boleslaw K. Szymanski at the Conference on Complex Systems.

Prof. Janusz Holyst presented today a talk on "Finding the source of local information cascades in complex networks," co-authored Krzysztof Suchecki, and Boleslaw K. Szymanski at the Conference on Complex Systems in Singapore. The talk discuss the well-known problem of finding a source of epidemic or information spreading process in complex network representing society is obviously of a great interest. In most realistic situations we know state of only few select observers, not all individuals. Like others, we consider each node in the network as potential source, assume treelike approximation of spreading paths and calculate probability distribution of times of infection of observers if given node was the source. Limiting the information to the infected observers we can evaluate likelihood of each node being source as the probability of obtaining exact observed times from distribution associated with that node. This however, would ignore information about where the infection did not arrive at all. We include this, by calculating probability that given source s, nodes were reached or not as observed. This probability is calculated using the same treelike approximation for spreading paths, probabilities to infect along each connection link, and probability to become resistant. Our results indicate that while the likelihood derived from times locates the source in global epidemic cascades, for local cascades the reach likelihood is mostly responsible for success.