Skip to Content
Texas State University

NSF awards Sasha Dong CRII award to improve optimization algorithms by using machine learning

Featured Faculty

Jayme Blaschke | May 8, 2020

dong headshot
Zhijie “Sasha” Dong, assistant professor in Ingram School of Engineering, was awarded a Computer and Information Science and Engineering (CISE) Research Initiation Initiative (CRII) grant by National Science Foundation.

Zhijie “Sasha” Dong, an assistant professor in the Ingram School of Engineering at Texas State University, has been awarded a Computer and Information Science and Engineering (CISE) Research Initiation Initiative (CRII) grant by the National Science Foundation. 

The NSF CRII award, also referred to as the "Mini CAREER Award," supports untenured junior faculty to build enough preliminary results to successfully apply for an NSF CAREER award. The two-year, $175,000 grant will support Dong’s proposal, “CRII: CIF: A Machine Learning-based Computational Framework for Large-Scale Stochastic Programming.”

Dong’s project will create a machine learning-based computational framework to solve large-scale random programming problems effectively and efficiently by integrating machine learning techniques into optimization algorithms. The project begins in June.

The difficulty of using existing commercial software to solve complex problems with multiple uncertainties, such as logistical supply chain issues with varying demand, inspired Dong to look at it from a different perspective. 

“There is one kind of algorithm, heuristic algorithms, that are based on experience,” Dong explained. “Those experiences can come from nature, like how ants look for food, or from physics. Those kinds of algorithms can help you find ‘good enough’ solutions in a very short time.”

Heuristic algorithms quickly reach approximate solutions by weighing available information at each decision point, making them an attractive alternative when more traditional problem-solving techniques fail. This fits in well within the context of machine learning: the algorithm is run once, the results checked, and the process is repeated after incorporating the updated information. 

“The work we’re proposing is to see whether we can take advantage of those machine learning techniques to help us improve our problem-solving,” Dong said. “We’re hoping we can learn more, so we can expedite the entire process. People usually use optimization algorithms to improve the machine model. Very few people are thinking the other direction, which is to use machine learning to improve the optimization algorithm.

“That is what I proposed in this NSF grant. It has very broad applications: supply chain issues, logistics, transportation, disaster response and so on,” said Dong, who is already collaborating with AMD, Sabre Airline Solutions and the San Antonio Food Bank to apply her approach to solving industrial problems. “We can see a lot of potential applications for the proposed machine learning phased framework. It’s not purely theoretical.”

For more information, contact University Communications:

Jayme Blaschke, 512-245-2555

Sandy Pantlik, 512-245-2922