Structure Discovery in Complex Dynamic Networks.

Abstract

Structure discovery in dynamic networks is one of the important emerging problems in machine learning. Modern applications in organizational science, neuroscience, etc. -- as well as myriad intelligent automation defense applications -- provide uncharted challenges. Nonlinear dimensionality reduction provides a compelling foundation for structure discovery in dynamic networks, but there are currently, to our knowledge, no theoretically justified methods providing formal statistical discovery of complex dynamic network structure. Successful methodologies require scalable nonlinear dimensionality reduction applied simultaneously to a collection of data matrices. Our research will expand the knowledge of the theoretical and mathematical underpinnings of machine learning systems for such applications, as well as produce methods and algorithms for principled statistical inference appropriate for structure discoveryin complex dynamic networks that can be fully automated. We will endeavor to produce principled and practical theory, methods, and code for discovering underlying dynamics in time series of networks that a (well-versed) practitioner will be able to use without the assistance of a network data scientist.

Document Details

Document Type
DoD Grant Award
Publication Date
Apr 11, 2024
Source ID
N000142412278

Entities

People

  • Carey E. Priebe

Organizations

  • Johns Hopkins University
  • Office of Naval Research
  • United States Navy

Tags

Fields of Study

  • Computer science

Readers

  • Defense Technology Research and Development.
  • Neural Network Machine Learning.

Technology Areas

  • AI & ML
  • AI & ML - Machine Learning Algorithms