FogML- Intelligence Orchestration over Heterogeneous, Dynamic, and Contested Fog Learning Environments

Abstract

Artificial intelligence and machine learning (AI-ML) are expected to revolutionize network systems throughout the commercial and defense sectors. However, popular approaches for distributing AI-ML over networks like federated learning (FL) are not well suited for fog computing settings, where many layers of network nodes may separate the edge devices and main servers. While recent work has considered FL optimization strategies, there is still a wide gap between these techniques and the practical requirements of fog systems. Much of this stems from fog dynamics, manifesting in three dimensions- (i) local datasets, i.e., measurement statistics evolving with environment shifts; (ii) node availability, i.e., variations in node abilities to participate in AI-ML processes; (iii) connection topologies, i.e., evolving abilities to form-maintain communication links. Federated learning (FL) has emerged as a popular approach here- to learn a target AI-ML model across all data in the network, (i) each device trains a local version of the model on its own collected data, and (ii) these local models are periodically aggregated by a server into a global model, with the process repeating iteratively.

Document Details

Document Type
DoD Grant Award
Publication Date
Feb 06, 2025
Source ID
FA95502410083

Entities

People

  • Christopher Brinton

Organizations

  • Air Force Office of Scientific Research
  • Purdue University
  • United States Air Force

Tags

Fields of Study

  • Computer science

Readers

  • Computer Networking
  • Economics
  • Neural Network Machine Learning.

Technology Areas

  • AI & ML
  • AI & ML - DoD AI Strategy
  • AI & ML - Neural Networks