Information Fusion from Heterogeneous Sources
Abstract
The project develops algorithmic foundations of information fusion for an autonomous learning agent to fuse information from other agents of heterogeneous tasks, models, and data. Specifically, the team will develop novel mathematical theories and computational algorithms for 1) fusing multiple already-learned models targeting small tasks into a powerful model targeting a unified task, 2) enhancing the reliability and identifiability of fusion, 3) reducing model complexity and sample size needed for training, 4) fusing models that have been learned from heterogeneous feature variables, and 5) enhance the prediction power with a large amount of unlabeled data. The team will demonstrate their work with data applications in various domains, including object classification, weather prediction, deep model compression, sensor monitoring, and decentralized learning. The developed research is envisioned to enhance the information fusion capability of general machine learning systems to extract actionable information more quickly and reliably.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 01, 2023
- Source ID
- W911NF2310315
Entities
People
- Jie Ding
Organizations
- Army Contracting Command
- United States Army
- University of Minnesota