Fusion And Inference From Multiple And Massive Disparate Distributed Dynamic Data Sets
Abstract
We have developed the first principled methodology for two-sample graph testing; designed a provably almost-surely perfect vertex clustering algorithm for block model graphs; proved analogues of classical limit theorems for the adjacency and Laplacian embeddings for random graphs, which have led, in turn, to significantly improved algorithms for latent position estimation; established the accuracy of and efficiently implemented a fast, successfully scalable program for an approximate solution to the NP-hard problem of matching graphs; developed efficient methods for vertex nomination in graphs; determined precisely how to mitigate information loss across shuffled networks. This has led to dozens of papers published in top journals. Moreover, we have employed these theoretically-justified techniques on a suite of applications, conducting end-to-end analyses of real data from domains as varied as neuroscience, speech and language processing, threat detection, and social networks.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 2017
- Accession Number
- AD1037242
Entities
People
- Carey E. Priebe
Organizations
- Johns Hopkins University