Thanks to two awards from the National Science Foundation, University of Utah Electrical and Computer Engineering Assistant Professor Cunxi Yu had been granted nearly $1.13 million to develop multidisciplinary research across formal methods and deep learning.
The first project led by Yu will focus on breaking down Boolean Networks with machine learning. Boolean networks, models mapping out variables and how they interact, are the best way for researchers to dive into complex dynamic behavior but have been rapidly growing. From millions of nodes to billions, the algorithms researchers once used to reason and optimize the networks are proving ineffective, making the networks less practical for real-world applications.
“When you have problems you want to solve those problems, right? That’s the main reason I started looking in this direction,” Yu said.
For the next three years, Yu and his team will leverage graph learning and neural networks to speed up large scale Boolean networks reasoning.
Because realistic deep learning applications typically have substantial computational and memory requirements there are some major challenges. The second project lead by Yu aims to employ formal methods to significantly improve the performance of DNN execution while providing useful quality guarantees that will enable wider deployment of deep learning. This project is led by the University of Utah in collaboration with Cornell University.
Yu’s final goal is to develop generic open-source frameworks with joint research efforts on formal methods and machine learning, which will be accessible to everyone.
“It can be used for any further development in any other domain, not just for me or for our research,” Yu said, “Whether it’s for bioinformatics or computer science, people can use this.”