Kaiwen LI

Student at BFSU

kevin@bfsu.edu.cn

[CV]

About Me

Hello! I'm a Master's student in Management Science & Engineering at Beijing Foreign Studies University (opens new window) where I am fortunate to be advised by Prof. Xi Chen (opens new window). My research interests include topics in large-scale sequential decision-making and statistically efficient optimization under uncertainty. Specifically, I am interested in:

Methodologies:

• Approximate Dynamic Programming / Reinforcement Learning
• (Distributionally) Robust Optimization
• Collaborative Learning

Applications:

• Transportation and Logistics
• Renewable Energy

Education

• International Business School, Beijing Foreign Studies University, China
  Master in Management Science & Engineering
  Rank: 1/24
  2022 - Present

• School of European Language and Culture, Beijing Foreign Studies University, China
  Bachelor of Arts in Italian Language and Literature
  2018 - 2022

Research

Publications

• Technician routing and scheduling with employees’ learning through implicit cross-training strategy
  Xi Chen, Kaiwen Li, Sidian Lin, Xiaosong Ding
  International Journal of Production Economics 271, 109208 (2024)
  [Link (opens new window)]
  Model: Vehicle Routing Problem (VRP) and Markov Decision Process (MDP)
  Algorithm: Approximate Dynamic Programming (ADP, closely related to RL)

• Gradient boosting decision tree in the prediction of NOx emission of waste incineration
  Xiaosong Ding, Chong Feng, Peiling Yu, Kaiwen Li, Xi Chen
  Energy, 126174 (2022)
  [Link (opens new window)]

Working Papers

• Data Provision via Federated Learning Platforms under Competition
  Joint work with Prof. Mingxi Zhu (opens new window)
  

• Near-Optimal Cost Function Approximation for Technician Routing and Scheduling
  Joint work with Prof. Xi Chen (opens new window)
  Presented at 2024 INFORMS Annual Meeting
  Brief Intro: Cost Function Approximation (CFA) can be written as: $x_t^{\text{CFA}} = \arg \min_{x\in X} \left( c(x_t^\pi, \xi_t) + \Omega(x_t^\pi, \xi_t)\right)$, where $x_t$ and $\xi_t$ refer to decision and exogenous random vector in stage $t$, $c(\cdot)$ is cost function and $\Omega(\cdot)$ is any parametric function with parameter $\lambda$.
  In CFA, how to design a “good” function $\Omega(\cdot)$ and choose a “good” parameter $\lambda$ is a fundamental problem, yet previous works usually propose specific formulations for specific problems and, to the best of my knowledge, there does not exist a general framework about how to design a “good” CFA.
  In this paper, we aim to develop near-optimal formulation of CFA utilizing state abstraction and the relationship between regularization and robust optimization.

• Measuring the robustness of international agricultural trade: A complex network approach
  Xi Chen, Kaiwen Li, Xiaosong Ding
  Under review at Chaos, Solitons and Fractals

Awards & Honors

• The Mathematical Contest in Modeling 2021
  Finalist Winner (Top 1%)

• First Class Scholarship, Beijing Foreign Studies University
  2022,2023

• Chinese National Scholarship (Top 2%)
  2024