Yu Awarded DARPA Grant

Assistant Professor Cunxi Yu has been awarded an 18-month, $500,000 grant from the US Defense Advanced Research Projects Agency (DARPA) for his proposal titled Structure-agnostic Combinatorial Optimizationn via Dataless Neural Architecture. His project aims to explore high-performance and intelligent methods for solving classic combinatorial optimization problems.

Through his research, Yu plans to address three challenges that are presented by classic combinatorial optimization (CO). These include an unfavorable speed-quality trade-off, ineffective optimization from data and proxy cost function, and a lack of hardware specialization opportunities. In order to address these challenges, Yu and his team will develop a novel dataless neural architecture framework for COs.

The goal of this research will be to discover solutions capable of solving these previous issues with COs. The desired results will encompass flexible and compact encoding of the solution space, ability to search with differentiable optimization while integrating learnable cost models, and hardware acceleration. This grant will contribute to the first effort in multi-disciplinary research towards dataless machine learning.

Yu joined the ECE Department in 2023. His research focuses on novel algorithms, systems and hardware design for computing and security.

Published February 6, 2025