Variational Methods on Graphs for Identification on Transaction Networks
Abstract
The project is fundamental unclassified research on graph detection methods applied to synthetic data, which was modeled on transaction data and provided to the project by the government. The project is critical to national security because such data is often the primary information known regarding adversaries interested in carrying out adversarial activities. The mathematical structure of the information to be analyzed will be a directed graph with a large number of nodes and even more edges. Additional information may be present including time stamps and meta-data however such information will not be the primary focus on the research. With the graph structure as the primary information, our goal is to detect subgraphs/activities on graphs of a specified known type but with information that potentially has noise, missing information and mis-alignment of the graph.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 16, 2022
- Accession Number
- AD1177299
Entities
People
- Andrea Bertozzi
Organizations
- University of California, Los Angeles