Jinyang Liu (刘锦阳)

3+5 PhD student

University of Copenhagen
Universitetsparken 5 (Department of Mathematical Sciences)
2100 Copenhagen

jl (at) math.ku.dk

About

– Born on 14. August 2002 in Shenzhen, China

I am currently a 3+5 PhD student in statistics and machine learning at the University of Copenhagen (UCPH). My research focuses on regression, intepretability, and tree-based methods such as gradient boosting and random forests. I hold a Bachelor’s degree in Mathematics from UCPH (June 2023), where I specialized in statistics and developed a strong foundation in measure and probability theory.

In addition, I am an experienced software engineer and have worked extensively with DevOps. I am proficient in Python and R, which I have used to develop microservices on Google Cloud.

I am developing a machine learning library for regression in Rust, focusing on creating an interpretable “glass-box” model as an alternative to traditional black-box methods.

Education

• University of Copenhagen, Sep 2023 - Aug 2028 (expected)
  3+5 PhD student in statistics and machine learning
  Advisor: Munir Hiabu
• University of Copenhagen, Sep 2020 - June 2023
  BSc in Mathematics
• H.C. Ørsted Gymnasiet i Lyngby, Aug 2017 - June 2020
  Danish technical high school education (HTX)

Research

Liu, J., Steensgaard, T., Wright, M. Pfister, N. Hiabu, M.
Fast Estimation of Partial Dependence Functions using Trees

arXiv preprint, 2024.

We provide a novel and fast method for computing partial dependence funtions for tree-based prediction models such as XGBoost and Random Forests. The implementation has since been integrated in the R-package glex.

CitearXivCode

@misc{liu2024fastestimationpartialdependence,
 title={Fast Estimation of Partial Dependence Functions using Trees},
 author={Jinyang Liu and Tessa Steensgaard and Marvin N. Wright and Niklas Pfister and Munir Hiabu},
 year={2024},
 eprint={2410.13448},
 archivePrefix={arXiv},
 primaryClass={cs.LG},
 url={https://arxiv.org/abs/2410.13448},
 }