Xuchen Gong

MS in Machine Learning Machine Learning Department, School of Computer Science Carnegie Mellon University Email | Google Scholar | GitHub

About me

Hello, I am a second-year MS student at CMU. Previously, I obtained my B.S. in Data Science from Duke Kunshan University and Duke University. I am studying the tradeoff between accuracy and other desirable properties of machine learning including adversarial robustness, robustness to distribution shifts, and parameter efficiency in multiple practical settings, such as cross-model transfer and federated learning. My investigations on trustworthy ML have drawn me to think about two directions of questions:

1. Trustworthy learning: Can we improve the tradeoff between in-distribution accuracy and metrics such as robustness, efficiency, fairness, and privacy, to make the algorithms more reliable to use in practice? How do these metrics interact with each other?

2. Deep representation learning: How can we better understand and leverage insights from the dynamics of deep learning to improve the above tradeoffs?

Research

* denotes equal contribution

• Unraveling the Complexities of Simplicity Bias: Mitigating and Amplifying Factors
  Xuchen Gong*, Tianwen Fu*
  NeurIPS 2023 Mathematics of Modern Machine Learning Workshop
  
  

• Cross-modal Assisted Training for Abnormal Event Recognition in Elevators
  Xinmeng Chen*, Xuchen Gong*, Ming Cheng, Qi Deng, and Ming Li
  ACM International Conference on Multimodal Interaction (ICMI), 2021
  
  

• An Abnormal Activity Detection Method and System Based on Video Monitoring
  Ming Li, Xinmeng Chen, Xuchen Gong, Ming Cheng, Yueran Pan, Qi Deng
  Patent CN113269111A 2021
  

Teaching

• COMPSCI 101 - Introduction to Computer Science
  Peer tutor, DKU Spring 2020, Spring 2021

• GLOCHALL 201 - Global Challenges in Science, Technology, and Health
  Peer tutor, DKU Fall 2019