Optimizing Student Success

“We are pleased to announce that your paper, “Natural Hypergradient Decent: Algorithm Design, Convergence Analysis and Parallel Implementation,” has been selected for the finalist of INFORMS Quality, Statistics, and Reliability (QSR) Best Student Paper Award, and the final competition will take place at the INFORMS Annual Meeting, where finalists will present their papers at the "QSR Best Student Paper Competition" session.

ISyE 3rd-year PhD student Deyi Kong received this exciting news in September 2025—for her first research paper, and her first time attending the INFORMS conference.

Out of 39 submissions for the Best Student Paper Award, and a rigorous double-blinded review process, Kong became one of four finalists to be selected for the award. Her paper was chosen alongside the other finalists from Massachusetts Institute of Technology, Georgia Tech, and University of Washington.

As a finalist, Kong explains that “a student paper has to have good insight and innovation. The idea of the paper must be relevant, meaningful, and have a practical impact on QSR applications. The work should have innovative ideas, theoretical justification and experimental research that proves what was developed actually works.”

Kong’s paper, written with assistance from her advisor Professor Shancong Mou, focuses on the development of an algorithm for unconstrained bi-level optimization, with the inner problem focusing on large scale learning problems. This algorithm can be applied to many industrial settings.

This problem appears in many engineering and machine learning applications,” Kong says. “In manufacturing, for example, we often want to perform automatic inspection of products. However, we may not have enough real data to train a supervised model, so synthetic data generation has become a popular way to mitigate this issue. But how do we evaluate the quality of the synthetic data? This algorithm can be used to qualify the dataset so that we can move forward with the manufacturing process more effectively.”

The INFORMS Annual Meeting, which took place from October 26-29, 2025 in Atlanta, Georgia, was where Kong presented her work in front of her fellow finalists and judges, as well as other attendees. This presentation, which was the competition for all of the finalists, was roughly 17 minutes long. The presentation was followed by a Q&A session from attendees and judges.

Though Kong did not win the award, she is extremely proud of the work she has done, as well as deeply grateful to Professor Mou, Professor Zaiwei Chen (Edwardson School of Industrial Engineering, Purdue University) and Professor Shuzhong Zhang.

“They gave me guidance and advice throughout this process, which contributed to making this finalist recognition possible. I also feel very fortunate to learn from them.”