In order to deepen the cultivation of top-notch innovative talents, strengthen the scientific research ability and innovative spirit of undergraduate students, and solidly promote discipline construction, the final defense of the 2026 undergraduate top-notch scientific research training program project of the School of Statistics and Data Science was successfully held on the afternoon of May 7th. After a year of systematic cultivation and in-depth research, the students of the 2023 Statistics Experimental Class have completed research results in the fields of high-dimensional statistics, machine learning, reinforcement learning, large model evaluation, bioinformatics, quantitative finance, etc., focusing on showcasing the effectiveness of academic training and innovative potential.
This defense invited teachers Han Miao, Qiu Yixuan, and Zhou Huijuan from the college to serve as defense judges, conducting comprehensive evaluations from the perspectives of topic value, innovative methods, rigorous experiments, achievement implementation, and academic standards. Each project team will report their research background, model construction, experimental verification, and application prospects in sequence, fully demonstrating their solid statistical theory foundation, proficient programming implementation ability, and scientific research literacy in facing real and complex problems.
This year's project is deeply cross integrated, closely related to real needs, and has outstanding practicality. It focuses on statistical theory and algorithm innovation, proposing new models and efficient solving methods around key issues such as high-dimensional feature selection, quantile regression, distributed reinforcement learning, optimal binning, and sparse variable selection. It also makes breakthroughs in real-world applications, forming interpretable and implementable statistical solutions in fields such as large model automatic evaluation, financial AI intelligent agent evaluation, genetic data feature selection, oral health dynamic intervention, PM2.5 pollution traceability warning, quantitative investment portfolio configuration, and vertical medical data modeling, while balancing theoretical rigor and practical value.
After strict review by the evaluation committee, a total of 12 projects have successfully passed the final acceptance, and multiple achievements have shown outstanding performance in theoretical innovation, algorithm efficiency, and application effects, fully reflecting the training orientation of the top-notch training program of "solid foundation, strong cross cutting, and emphasis on practice". Among them, "A Unified Framework for Simulated Optimal Binning and Sparse Variable Selection" (project members: Li Yuge, Lu Hongbo, Chen Zifan, instructor: Feng Xingdong), "Graph LassoNet: LassoNet Model and Its Application Integrating Turalaplus Regularization" (project member: Yu Yue, instructor: Wu Mengyun), and "Single Indicator Threshold Model and Treatment Sensitive Subgroup Identification Method Based on Longitudinal Data" (project member: Shan Hanqi, instructor: Han Miao) were awarded excellent projects.
The undergraduate top-notch research training program is the core carrier for the college to implement the principle of "people-oriented" and create high-quality undergraduate education. It is also a key platform for connecting "curriculum learning research training innovative practice". In the future, the School of Statistics and Data Science will continue to optimize its top-notch talent training system, continuously improve its undergraduate research training platform and practical innovation platform, produce more high-level achievements, and provide the country with more composite data science talents who possess solid statistical skills and strong innovation capabilities.
Contribution | Xiong Jifeng
Image provided | Lu Hongbo (Student)
Review | Li Tao
