Optimal Targeting in Dynamic Systems
Abstract
Approved for Public ReleaseMachine learning based methods for optimal targeting have been rapidly adopted across several fields. A limitation of these methods, however, is that they have largely been developed for optimizing static, one-time decisions. In dynamicdecision-making problems, any action we take may induce both an immediate benefit and a system change with indirect, longer-term consequences. The goal of the proposed research is to extend the machine learning toolkit for optimal targeting to the dynamic setting, and to develop methods that enable planners to optimize decisions by rigorously trading off short- and long-terms benefits of targeted actions.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 24, 2024
- Source ID
- N000142412091
Entities
People
- Stefan Wager
Organizations
- Office of Naval Research
- Stanford University
- United States Navy