My research interests focus on learning theory now, i.e. bandit and reinforcement learning theory. I'm also interested in theoretical statistics. If you are also interested in this field and would like to communicate with me, feel free to contact me.

Because I majored in mathematics and learned a lot of mathematics courses, I tend to do pure theory during my undergraduate study. But I am personally interested in all kinds of directions and hope to try more directions in the future to solve really interesting and meaningful problems.

I am working on projects related to bandit and RL with Professor Xu from Duke university, and here are some notes on the papers I have read so far.

Research Interests

  • Artificial Intelligence
  • Reinforcement Learning, Bandit
  • Information Theory
  • Data Science
  • Optimization
  • High Dimensional Statistics

Research Experience

Optimal Batched Linear Bandits
Aug. 2023 – Present
Advisor: Prof. Pan Xu
Duke University, Department of Electrical and Computer Engineering
  • Inspired by extensive literature review, devised a groundbreaking algorithm that strives for asymptotic and finite-time optimality in the linear bandits setting, an achievement previously unattained.
  • Further adapted the algorithm into a batched version, addressing common real-world problems.
  • Confirmed our algorithm's superiority over existing methods by conducting rigorous experimentation, showcasing its practical efficacy in linear bandits.
Bandit Theory Reading Group
Jan. 2023 – Present
Advisor: Prof. Pan Xu
Duke University, Department of Electrical and Computer Engineering
  • Actively participated in ongoing discussions and consistently stayed abreast of the latest advancements in Bandit theory problems through comprehensive paper readings.
  • Delivered seven presentations as the primary presenter, introducing the concepts of Information-Directed Methods, including the Information-Theoretic Thompson Sampling (TS) analysis approach and the Information-Directed Sampling Algorithms.
  • Took a leading role in guiding discussions and actively stimulated research project ideas during these presentations, fostering a collaborative and innovative environment.
Reinforcement Learning Theory Reading Group
June 2023 – Aug. 2023
Advisor: Prof. Pan Xu
Duke University, Department of Electrical and Computer Engineering
  • Engaged proactively within a specialized reinforcement learning theory-focused reading group, actively participating in intricate discussions and analyzing state-of-the-art research papers.
  • Collaboratively tackled complex theoretical challenges and explored innovative solutions within the group, leading to a deeper grasp of reinforcement learning principles.
  • Assumed a leadership role by spearheading discussions on Strategic Exploration in MDPs, encompassing advanced concepts like linearly parameterized MDPs and Generalization with Bounded Bellman Rank, during a comprehensive book study on Reinforcement Learning theory.
Optimal Production Planning and Water Saving Decision under Sticky Prices
Undergraduate Innovation and Entrepreneurship Training Program (Download thesis in Chinese)
Jun. 2022 – May 2023
Advisor: Prof. Lijun Bo
USTC, Department of Mathematical Sciences
  • Developed a company production and water rights trading model with sticky prices, addressing slow price adjustments despite market dynamics.
  • Derived Hamilton-Jacobi-Bellman equation solutions, yielding optimal feedback strategies for production planning and water rights trading decisions.
  • Conducted numerical simulations, analyzing value function structural properties and enhancing decision-making through parameter sensitivity analysis.
  • Uncovered valuable insights for companies, providing actionable recommendations for informed production planning and water rights trading strategies.
Recursive Partitioning and Applications
Sept. 2022 – Jan. 2023
Advisor: Prof. Bo Zhang
USTC, Department of Finance and Statistics
  • Led an in-depth exploration of recursive partitioning methods in machine learning during a semester-long seminar. Conducted extensive literature reviews to gain comprehensive knowledge of the field.
  • Over the course of the seminar, assumed the role of the primary presenter, delivering ongoing presentations that elucidated the intricacies of Tree-Based Recursive Partitioning methods.
  • Pioneered experimental integration of hypothesis testing using Tree-Based p-Value into random forest analysis. Introduced an effective tree node partitioning criterion based on statistically significant inter-node differences, leading to a notable reduction in random forest tree partitioning error.
Model Selection in Linear Regression Based on Coefficient of Determination
Jun. 2022 – Jul. 2022
Advisor: Prof. Bo Zhang
USTC, Department of Finance and Statistics
  • Developed an innovative approach for variable selection problem in linear regression models. Quantified the impact of independent variables on explained variable variance and identified influential factors.
  • Employed a novel estimation method for explained variance that transcends the normality assumption and covariate sparsity constraints. Leveraged insights from estimating equation techniques used in high-dimensional linear models.
  • Validated the algorithm's efficacy through extensive simulations on synthetic datasets, showcasing its robustness and practical application in various scenarios.
High Dimensional Statistics Reading Group
Nov. 2022 – Present
Advisor: Prof. Bo Zhang
USTC, Department of Finance and Statistics
  • Initiated in-depth exploration by conducting literature surveys on high-dimensional statistics, actively participated in seminars to elucidate complex concepts to peers and instructors.
  • Collaboratively conducted literature surveys on statistical theory articles related to tensors, gaining a comprehensive understanding of this tool and sparking explorations for potential project ideas.
  • Conducted presentations on articles related to change-point problems and Gaussian mixture models, contributing to the dissemination of knowledge and facilitating discussions in the academic community.