Title: Spectra of Random Graphs II: Outliers
Speaker: Ioana Dumitriu
Speaker Info: University of California, San Diego
Brief Description:
Special Note:
Abstract:
Outliers in the spectra of random graphs are crucial to many machine learning applications (clustering, matrix completion, etc.) and, especially in sparse and inhomogeneous cases, their existence is much harder to determine than in random matrix theory (where we have the famous Bai-Yin 4th moment theorem). Methods of approach include a generalized method of high moments, often in conjunction with the very promising non-backtracking operator. We will explain the basics of these methods, and show how they work in the case of a few graph models.Date: Thursday, October 20, 2022