Liyuan Hu

Liyuan Hu

Ph.D. candidate in Statistics

Liyuan Hu

I’m a final year PhD student from London School of Economics and Political Science. My research interest includes reinforcement learning applications. I am very fortunate to be advised by Prof. Chengchun Shi from London School of Economics and Political Science.

Education

London School of Economics and Political Science | 2022.09 - 2026
Ph.D. in Statistics | Reinforcement Learning Track

  • Honor: Full Ph.D. Scholarship

Sun Yat-sen University | 2018.09 - 2022.06
B.S. in Statistics, School of Mathematics

  • GPA: 4.3/5.0 (Ranked 2/78)
  • Relevant Coursework: Statistical Learning, Complex Data Analysis, Data Structures, Mathematical Statistics
  • Honors: National Scholarship (2×), First-Class University Scholarship (3×), National Second Prize in Chinese Mathematical Contest in Modeling (2019)

Professional Experience

TikTok Global Monetization Product and Technology | July 2025 - Present
Algorithm Engineer, Commerce Ads Technology

  • Participated in TikTok e-commerce platform GMV Max product development, conducting automated development, data analysis, and A/B testing to improve overall ROI revenue

Huatai Securities Research Institute | April 2025 - July 2025
Alpha Team Researcher, Financial Engineering Group

  • Enhanced value factors using large language models based on annual report data
  • Balanced context length limitations, cost constraints, and training effectiveness in machine learning methods for LLM training
  • Implemented end-to-end factor mining using Graph Neural Networks to model cross-sectional stock return factors

Invesco Great Wall Fund | October 2024 - February 2025
Quantitative Researcher, Quantitative and Index Investment Department

  • Developed end-to-end index enhancement strategies
  • Reproduced and improved LinSAT (a differentiable combinatorial optimization neural network component), optimizing training time by nearly 10× while maintaining equivalent performance
  • Achieved 10%-40% improvement in Information Ratio (IR) for enhancement strategies on CSI 300, CSI 500, and CSI 1000 indices compared to traditional non-end-to-end multi-factor stock selection frameworks

Research Projects

Q-Function Strategy Optimization Addressing Inter-Group Data Correlation | January 2023 - May 2025
First Author

  • Investigated applications of Generalized Estimating Equations in reinforcement learning
  • Proposed a novel Fitted Q-iteration algorithm that improves learning strategy effectiveness by estimating inter-group data correlations
  • Submitted to Statistics Journal

Deterministic Linear Reinforcement Learning Strategy Optimization | August 2023 - Present
First Author

  • Developed linear deterministic reinforcement learning strategies suitable for device-constrained environments
  • Addressed existing device limitations in storage and design aspects
  • Conducted simulation validation on medical school simulators

Strategy Optimization for Non-Stationary Heterogeneous Data | April 2022 - February 2025
First Author

  • Developed novel reinforcement learning algorithms for temporally non-stationary and individually heterogeneous data
  • Enhanced reinforcement learning applicability and efficiency in dynamic environments
  • Preparing submission to Journal of the Royal Statistical Society Series B

COVID-19 County-Level Mortality Risk Analysis in the United States | April 2020 - September 2021
First Author

  • Conducted risk analysis of COVID-19 mortality rates across 3,125 U.S. counties
  • Explored health and socioeconomic factors related to mortality rates
  • Published in Infectious Diseases of Poverty

Software Development

abess: Fast Best Subset Selection Package (PyPI & R CRAN) | December 2020 - September 2021
First Author

  • Co-developed the abess library, implementing and extending core algorithms based on C++ kernel
  • Developed corresponding R interface for the library
  • Created efficient toolkit for best subset selection problems in machine learning (linear regression, classification, PCA)
  • Achieved 20× speed improvement compared to existing tools
  • Published in The Journal of Machine Learning Research

bestridge: Best Subset Selection with Ridge Penalty Package (R CRAN) | February 2020 - March 2021

  • Responsible for algorithm design and C++ kernel implementation
  • Led R interface development

Technical Skills

  • Programming Languages: Python, C++, R
  • Specializations: Machine Learning, Reinforcement Learning, Quantitative Finance, Statistical Modeling
  • Languages: English (TOEFL 108), Chinese (Native), CET-6

Publications & Achievements

  • Published research in Infectious Diseases of Poverty and The Journal of Machine Learning Research
  • Multiple submissions to top-tier statistics journals in progress
  • National-level competition recognition in mathematical modeling
  • Consistent academic excellence with multiple scholarship awards