Weiqin is a final-year PhD candidate in ECSE whose research interests lie in reinforcement learning (RL), optimization and foundation models. RL is a trial-and-error learning that allows an AI agent to learn from experiences. It is very exciting and promising that RL has been developed to address a wide variety of decision-making problems, such as strategy games, robotics, recommender systems, and large language model post-training.
As a freshman at North China Electric Power University (NCEPU), Weiqin was enamored with the mathematical theories and algorithms behind applications, choosing to major in Information and Computing Sciences. Gradually he became less satisfied with theoretical study and pivoted to industrial applications. Ranking number 2 in his major gave him the flexibility to switch to Energy and Power Engineering, a highly competitive major at the University due to the rigorous and comprehensive curricula. Weiqin was very busy during his undergraduate years, participating in multiple research projects spanning Mathematics and Engineering. These included a nonlinear wave solution and control project, the Lattice Boltzmann model for high dimensional nonlinear partial differential equations (NPDEs), and the simulation of the thermal behavior of the low-temperature phase change material (PCM), which led to two first and second author publications. Additionally, he earned a meritorious winner prize in 2019, awarded to the top 7% of participants globally in the Interdisciplinary Contest in Modeling organized by Consortium for Mathematics and Its Application of the United States. His interest in mechanical manufacturing led to possession of a utility model patent for a pumping irrigation device that uses river power. Weiqin attended Wrocław University of Science and Technology (WUST) in Poland for a one-year exchange, awarded to the very top students at NCEPU. While earning perfect scores in the courses, he took the opportunity to conduct a project about the heat transfer of PCM and did a comparative study on the experimental and numerical melting process of the paraffin wax based on Computational Fluid Dynamics (CFD). Weiqin proposed that a scanning electron microscope could reveal the internal structure change of the PCM, confirming its homogeneous structure an lack of internal changes during phase transformations at characteristic temperatures. This novel idea led to his second author publication and fortified his interest in simulation and experimentation in engineering. Graduating in the top 5% of his university in a class of approximately 5 thousand, Weiqin earned the Outstanding Graduate Award. He also earned Third Prize as project leader in the First National Undergraduate Competition on Renewable Energy Technology, Renewable Energy Society of China. Weiqin was identified as a “Top Ten Technology Pacesetter,” an honor bestowed on one out of 1,250 students! Multiple honors and scholarships followed as Weiqin graduated a National Excellence Winner in the Student Innovation and Entrepreneurship Training Program.
Currently, Weiqin contributes to Professor Santiago Paternain’s lab and continues to be productive at the intersection of machine learning and dynamical systems. He has published 10 research papers since his start of PhD journey, including in top conferences and top journals. He was the recipient of the 2023 Belsky Award for Computational Sciences and Engineering by the RPI School of Engineering and has done two research internships at IBM T. J. Watson Research Center.
Though RL has shown great potential in solving many problems, its implementation has been limited to a few domains as it is besieged by several fundamental challenges, of which safety, poor generalization, and low sample efficiency are paramount. Thus the goal of Weiqin’s PhD research program is to develop algorithms to systematically break these severe barriers, which hinder the adoption and deployment of RL techniques in real-world applications. Over the past few years, Weiqin has completed several RL projects including safe RL, in-context RL, offline RL, control-based RL and currently RL for LLM reasoning and LLM safety alignment.
During his free time, Weiqin keeps active, playing ping-pong, basketball, badminton, and tennis.
