Estimating a Network from Multiple Noisy Realizations

• Activity Number: 330 - Advances in Inference of Networks
• Type: Topic Contributed
• Date/Time: Tuesday, August 1, 2017 : 10:30 AM to 12:20 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Can Le* and Liza Levina
• Companies: University of California Davis and University of Michigan
• Keywords: Network ; Noisy network ; Community detection ; EM algorithm ; Link prediction

Abstract

Complex interactions between entities may be mathematically represented as a network using a set of edges between them. In practice, the network is often constructed from various measurements and inevitably incurs some errors. In this talk we consider the problem of estimating a network from multiple noisy observations where edges of the original network are observed with certain false positive and false negative noises. We assume that the original network contains a community structure and within each community edges are observed with similar errors. This allows us to derive an efficient and theoretically guaranteed method to estimate the noises and in turn the original network. We show that the performance of our method on simulated data is similar to an optimal method when all parameters are assumed to be known.