Structured Regression in Complex Networks by Fusion of Qualitative Knowledge and Big Data
Abstract
Title: Structured Regression in Complex Networks by Fusion of Qualitative Knowledge and Big Data Principal Investigator: Zoran Obradovic The scope of this project is developing a synergetic platform for predictive modeling of complex dynamic networks that will connect cutting-edge predictive machine learning models and rigorous mathematical methods for network reduction. The objective will be to unify data-driven and knowledge-driven approaches for structured network-based predictive analytics in order to broaden the scope of possible applications. The approach will consist of extending our structured regression models to explore in detail the hypothesis that a unified approach of integrating big data with high-level knowledge (ontologies, domain-based constraints etc.) is beneficial for predictive modeling of complex phenomena. The first task will be developing a framework for fusion of data, context, and qualitative knowledge for modeling dynamics in complex networks. This will be achieved by adding convex or concave penalty terms in the optimization function corresponding to multiple constraints from qualitative knowledge. The second task of scaling up the proposed solution to big applications will be achieved by compression of large complex networks for more efficient structured regression. The proposed framework is expected to allow for improving the performance of Navy sensing systems by providing more accurate predictions in dynamically changing environments through structured regression based on integration of observed data, domain knowledge, and context. Methods developed in this project will enable more complete decision support by allowing predictive modeling of large and complex evolving networks reduced in a way that maintains predictive information.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 12, 2016
- Source ID
- N000141512729
Entities
People
- Zoran Obradović
Organizations
- Office of Naval Research
- Temple University
- United States Navy