|
Abstract:
|
In connectomics, the study of the networked nature of the brain, one frequently is confronted with a collection networks corresponding to different individuals, scans, treatments or populations. Both as a summary statistic and for an input to subsequent analysis, an accurate estimate of the mean network for the population is necessary. In this talk we will investigate alternative methods to simple averaging of adjacency matrices including low-rank and robust methods. We study their performance in random graph models such as stochastic blockmodels as well as for neuroscience data.
|
