A distributed neural filter for finding depth-K skyline friends in social networks

Finding similar users in a social network is an essential element of personalized recommendation systems. Most traditional algorithms of this kind consider all the dimensions of user data as a whole and use this merged information to search for similar users. However, such a method has its flaws, as user information in various dimensions is generally independent. Recently multi-criteria skyline queries have been applied to search for similar users. However, these applications are only suited to centralized environments, which is not a common type of environment among social networks. Hence, this paper introduces a distributed neural filter to search for similar users in a distributed environment. By using the proposed filter, we can enhance the speed of identifying whether a user is similar to you in the distributed environment. That is, the operating speed of recommendation systems can be improved. Simulation results demonstrate the effectiveness and efficiency of the proposed identifier.