Federal Project Grant Award Summary

The University of Virginia received a $170,000 Project Grant from the National Science Foundation (NSF) Division of Mathematical Sciences under the Mathematical and Physical Sciences program (CFDA 47.049), awarded September 1, 2026, with completion targeted for August 31, 2029. The project develops statistical methods and scalable algorithms for extracting weak signals from large, noisy datasets without requiring prior knowledge of noise structure. The research introduces a novel framework based on cycle count statistics to address inference challenges in spiked models—where observed data comprises a low-rank signal matrix corrupted by heterogeneous noise—particularly in regimes where signal strength is comparable to noise levels and standard spectral methods prove ineffective.

The project deliverables include theoretical advances in statistical methods for complex data structures, computational algorithms designed for large-scale applications, and improved analytical approaches applicable to biomedical and genomic data analysis, social network analysis, and public policy design. Beyond research outputs, the award supports workforce development activities including graduate student mentorship, curriculum integration of modern statistical methods, and training activities such as a short course for early-career scholars. The work is performed at the University of Virginia's Charlottesville, Virginia campus.