Learning and Inferring Networks from Incomplete Data

Abstract

In an increasingly interconnected world, information, goods, and even diseases propagate across explicit and implicit networks. Depending on the application, network edges can be used to represent different types of connections, from physical wires between computers to friendships among people. And in many situations, one can hope to learn the connections or other properties of a target network by examining whatever outputs of the network are observable, or by smartly intervening in the network and analyzing the results of the intervention. Designing algorithms for learning such networks is the goal of the proposed research. The army applications of this area are numerous and significant. Some examples of problems broadly covered by this proposal include reconstructing an adversarys networks from intercepted or publicly available communications, better understanding supply networks from their disruptions, or even discovering hidden influence structures after observing the votes or writings of politicians or other actors.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 27, 2017
Accession Number
AD1058732

Entities

People

  • Lev Reyzin

Organizations

  • University of Illinois at Chicago

Tags

Communities of Interest

  • Human Systems

DTIC Thesaurus Topics

  • Abstracts
  • Agreements
  • Algorithms
  • Autonomous Agents
  • Contractors
  • Department Of Defense
  • Education
  • Engineering
  • Information Operations
  • Learning
  • Mathematics
  • Military Research
  • Multiagent Systems
  • Probability
  • Social Media
  • Social Networks
  • Students

Fields of Study

  • Computer science

Readers

  • Applied Combinatorial Optimization and Logic Circuit Design.
  • Political Violence and Terrorism Studies.
  • Regression Analysis.

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks