Robust Information Fusion on Social Networks

This paper introduces the general framework of statistical information fusion on social network, which plays an important role on understanding the human interaction and cooperation in the society. On many existing information platforms on the social network, people's relationship affects their decision on both static and dynamic way. We consider the scenario that agents refer others decisions which are made earlier as an information for taking any action. Thus how agents on the information platform connect to each other has great impact on the information fusion. With the help of percolation theory, we provide a minimax robust decision scheme for information fusion due to often the lack of complete information of social network structure. The simulations demonstrate that our proposed information fusion rule retains fine performance compared with classical one on different network structures, and thus great potential for social network applications.