DARPA Defense Sciences Office is seeking proposals for basic or applied reseach concepts in mathematical methods to systematically discover useful mathematical transformations – transformations of a system’s variables and parameters to representations that make solving complex modeling problems easier, faster, and potentially more interpretable. transformations of a system’s variables and parameters to representations that make solving complex modeling problems easier, faster, and potentially more interpretable. New approaches based on advances in scientific machine learning (SciML) that are computationally efficient, interpretable, generalizable, and ultimately useful are of particular interest.
This program will be comprised of 2 sequential project phases:
- Phase 1 (base) – 12 months – up to USD$500,000 – create a useful mathematical transformation, understand the limits of the validity of their transformation, and demonstrate their tools’ ability to generalize within their system class as compared to a SotA baseline
- Phase 2 (option) – 12 months – up to USD$750,000 – create a useful transformation for a supplied new instance in their chosen class; subsequently extend the tools to generalize across a second system class in an application across multiple instances; demonstrate ability to detect and mitigate the possible singularities arising as limits of transformation validity
Abstracts (encouraged but not mandatory) are due 24 March 2025, with full proposals due 2 May 2025.
This Disruption Opportunity is issued under the Program Announcement for Disruptioneering – DARPA-PA-24-04.
Read more
Documents
Disruption Opportunity – TRS: DARPA-PA-24-04-07
Program Announcement for Disruptioneering: DARPA-PA-24-04
