Jiajia Yu 余佳珈

Phillip Griffiths Assistant Research Professor in Department of Mathematics Duke University Office: 120 Science Drive, 222 Physics Building, Durham, NC, 27708 Email: jiajia dot yu at duke dot edu Google scholar   ORCID   Github CV

Research Interests

Overall, I am interested in applied and computational math.
Currently, I work on numerical methods and applications of mean-field games and optimal transport.

Employment

• 2023.08-2026.07 Duke University, Department of Mathematics , Durham, NC
  Phillip Griffiths Assistant Research Professor
  Mentors: Prof. Hongkai Zhao and Prof. Xiuyuan Cheng. I also work closely with Prof. Jian-Guo Liu.

Education

• 2018-2023 Rensselaer Polytechnic Institute, Department of Mathematical Sciences, Troy, NY
  Ph.D. in Mathematics
  Advisor: Prof. Rongjie Lai

• 2013-2017 Beijing Normal University, School of Mathematical Sciences, Beijing, China
  B.S in Mathematics and Applied Mathematics

Recent Preprints/Publications

• [2025.07] Our draft Joint Inference of Trajectory and Obstacle in Mean-Field Games via Bilevel Optimization is online.

• [2025.06] Our draft Equilibrium Correction Iteration for A Class of Mean-Field Game Inverse Problem is online. In this paper, we propose an iterative algorithm and two variants for reconstruct the ambient potential in mean-field games. The algorithm is different from generic optimization based method and exploits the structure of mean-field games. We also explore the connection between the potential identification problem in mean-field games and in linear parabolic equations.

• [2024.11] Our draft Convergence Analysis and Acceleration of Fictitious Play for General Mean-Field Games via the Best Response is online. In this paper, we give the convergence rate of fictitious play for non-potential mean-field games by studying the stability of the best response mapping.

• [2024.10] Our draft Zeroth-order Stochastic Cubic Newton Method with Low-rank Hessian Estimation is online.

• [2024.08] Our paper A bilevel optimization method for inverse mean-field games is accepted by Inverse Problems.

Upcoming Events/Talks

• [2025.07] IPAM Workshop on Sampling, Inference, and Data-Driven Physical Modeling in Scientific Machine Learning, UCLA, Los Angeles, CA
  Invited speaker, Learning and inference in mean-field games