Federal Project Grant Award Summary

The University of Florida Division of Sponsored Research received a $197,273 Project Grant awarded October 1, 2025, through the National Science Foundation's Division of Information and Intelligent Systems under the Computer and Information Science and Engineering program (CFDA 47.070). This collaborative research initiative will develop formal methods for synthesizing and verifying neural networks deployed on in-memory computing (IMC) systems through August 31, 2027. The project delivers algorithms and verification tools that enable robust artificial intelligence (AI) model deployment on emerging analog and hybrid analog-digital IMC hardware, which presents greater susceptibility to errors than traditional complementary metal-oxide-semiconductor (CMOS) technologies. Key deliverables include formal verification methods for neural networks accelerated using analog IMC, synthesis techniques for hybrid analog-digital IMC circuits using decision diagrams and machine learning, and fabricated in-memory computing system demonstrations.

The project addresses the critical need for reliable AI training with reduced power consumption as neural networks become increasingly prevalent. By combining expertise in formal methods, machine learning, computer-aided design, and hardware fabrication, the research synthesizes crossbars for neural networks and verifies bidirectional digital IMC circuits prior to physical demonstration. This effort contributes to advancing formal methods for real-world challenges in emerging computing systems while enhancing neural network reliability on in-memory circuits and fostering interdisciplinary collaboration across formal methods, machine learning, and hardware design disciplines.