Law of Large Graphs

Abstract

Estimating the mean of a population of graphs based on a sample is a core problem in network science. Often, this problem is especially difficult because the sample or cohort size is relatively small as compared to the number of parameters to estimate. While using the element-wise sample mean of the adjacency matrices is a common approach, this method does not exploit any underlying graph structure. We propose using a low-rank method together with tools for dimension selection and diagonal augmentation to improve performance over the naive methodology for small sample sizes. Theoretical results for the stochastic blockmodel show that this method will offer major improvements when there are many vertices. Similarly, in analyzing human connectome data, we demonstrate that the low-rank methods outperform the standard sample mean for many settings. These results indicate that low-rank methods should be a key part of the tool box for researchers studying populations of graphs.