A0709
Title: A local perspective in general latent space network models
Authors: Rachel Wang - University of Sydney (Australia) [presenting]
Lijia Wang - City University of Hong Kong (Hong Kong)
Xin Tong - University of Southern California (United States)
Xiao Han - University of Science and Technology of China (China)
Abstract: The neighborhood effect in social networks significantly influences an individual's decision-making process, opinion formation, and various other personal dynamics. Thus, to understand the role of social networks in shaping individual behaviors and attitudes, it is important to begin with an understanding of an individual's localized viewpoint within the global network context. A general latent space network model and the problem of inferring the latent positions of the nodes are considered, utilizing only a partial information network centered on a given individual. Using a projected gradient descent algorithm, the convergence rate of the estimates is shown to depend on the neighborhood features of the node and a quantity is defined accordingly to measure the amount of bias in this individual's local view. Using simulated and real networks, particularly the co-sponsorship network in the US Congress, these local estimates of latent positions are compared with the global ones and show how the framework allows us to obtain a more nuanced understanding of local perspectives within social networks.
