Research & Innovation

Zhuo Feng Receives $550,000 NSF Grant to Explore Graph Learning Techniques

Feng's research plan will potentially advance the state of the art in spectral graph theory and electronic design automation (EDA)

circuit illustration

Principal Investigator Zhuo Feng, associate professor in the Department of Electrical and Computer Engineering at Stevens Institute of Technology, was recently awarded a grant of $550,000 from the National Science Foundation for his project “SHF: Small: Learning Circuit Networks from Measurements.”

Zhuo Feng

Graph learning techniques have shown promising results for various important applications such as vertex (data) classification, link prediction (recommendation systems), community detection, drug discovery, partial differential equation (PDE) solvers, and electronic design automation (EDA). However, the state-of-the-art graph topology learning methods do not scale to large data sets due to their high algorithm complexity. Existing graph topology learning methods cannot be efficiently applied in EDA tasks considering the sheer size of modern integrated circuit systems that typically involve millions or billions of elements.

Feng's research plan will investigate the first scalable yet sample-efficient spectral method for learning graphs (circuit networks) from a few voltage and current measurements. He plans to investigate the important EDA applications that include data-driven, physics-informed PDE solvers for much faster multiphysics simulations; and data-driven vectorless power/thermal integrity verification for worst-case performance estimation using a few voltage/current measurements.

The success of this research plan will immediately allow the development of scalable data-driven, physics-informed EDA algorithms for modeling, simulation, optimization, and verification of integrated circuits (ICs). The accomplished theoretical results will potentially advance the state of the art in spectral graph theory, scientific computation, as well as graph-based machine learning (ML). The practically efficient methods developed through this research will open avenues for developing data-driven, physics-informed EDA applications.

Feng said, “New spectral algorithms for estimating probabilistic graphical models will open avenues for developing highly efficient data-driven electronic design automation (EDA) algorithms.”