Federal Grant Award Summary

Rutgers, The State University received an $850,000 Project Grant from the National Science Foundation (NSF) Division of Information and Intelligent Systems under the Computer and Information Science and Engineering program (CFDA 47.070), effective July 1, 2026 through June 30, 2030. The award funds research to develop an output-sensitive algorithm for efficient convex hull computation in high dimensions, addressing a critical computational bottleneck in formal verification of deep neural networks. Unlike existing methods whose complexity scales with both the number of input points and dimensions, the project's novel algorithmic approach reduces computational complexity to depend only on the number of vertices on the convex hull itself, enabling more scalable verification of neural networks in safety-critical applications.

The research deliverables include foundational advances in convex hull computation algorithms with direct applications to neural network verification and the generation of correctly-rounded mathematical libraries. The project will also advance educational outcomes through instruction of graduate and undergraduate students on foundational computing abstractions. By enabling more accurate and efficient formal verification techniques for deep learning systems, this research addresses the growing need for robustness and safety assurance as neural networks expand into safety-critical domains.