Yifang Chen 陈一方

Ph.D student, Paul G. Allen School of Computer Science & Engineering, University of Washington Email: yifangc at cs dot washington dot edu Google Scholar Instagram Twitter

About me

I am currently a forth year Ph.D student in computer science and engineering from University of Washington. I am very fortunate to be advised by Prof. Kevin Jamieson and Prof. Simon Shaolei Du.

My general research interests lie in designing practical and adaptive machine learning algorithms with strong theoretical guarantees. One of my main focus is on corrupted and non-stationary online decision making setting. Recently I am growing my interest in applying active learning and experimental design techniques in large-scale models (Computer Vision, LLM, multi-modal) to improve the training and label efficiency. Particularly, from the empirical side, I am interested in doing a through investigation on how modern pre-trained models can help downstream data collection and how the human-in-the-loop algorithm can further improve the large scale models. From the theoretical side, I am interested in designing online/active learning algorithm beyond the linear setting, and combine that with the deep representation learning theory.

Check out our website LabelTrain.ai, which is an ongoing open-source project for experimentally exploring the strengths and weaknesses of label-efficient learning algorithms

Prior to starting my PhD study, I did my master and undergrad in electrical engineering from University of Southern California advised by Prof. Haipeng Luo. I want to specially thank him as well as my colleague Chen-Yu Wei, who led me into the world of learning theory from nowhere.

Research

My detailed research interests include

• Active learning and its application in fine-tuning

• Experimental design and its application in representation learning

• Online learning, bandits, Reinforcement learning theory

• Experimental design for scientific applications

Selected publications (reverse chronological)

1. [Preprint] Variance Alignment Score: A Simple But Tough-to-Beat Data Selection Method for Multimodal Contrastive Learning.
   Yiping Wang*, Yifang Chen*, Wendan Yan, Simon Du, Kevin Jamieson.
   

2. [Preprint] Experimental Design Framework for Label-Efficient Supervised Finetuning of Large Language Models.
   Gantavya Bhatt (equal), Yifang Chen (equal), Arnav M. Das (equal), Jifan Zhang (equal), Sang T. Truong, Stephen Mussmann, Yinglun Zhu, Jeffrey Bilmes, Simon S. Du, Kevin Jamieson, Jordan T. Ash, Robert D. Nowak
   

3. [Journal of DMLR][Github] LabelBench: A Comprehensive Framework for Benchmarking Label-Efficient Learning.
   Jifan Zhang (equal), Yifang Chen (equal), Gregory Canal, Stephen Mussmann, Yinglun Zhu, Simon Shaolei Du, Kevin Jamieson, Robert D Nowak
   

4. [ICML22] Active Multi-Task Representation Learning.
   Yifang Chen, Simon Du, Kevin Jamieson.
   

5. [ICML22] First-Order Regret in Reinforcement Learning with Linear Function Approximation: A Robust Estimation Approach. (Long presentation)
   Andrew Wagenmaker, Yifang Chen, Max Simchowitz, Simon S. Du, Kevin Jamieson
   

6. [ICML’21] Improved corruption robust algorithms for episodic reinforcement learning.
   Yifang Chen, Simon Du, Kevin Jamieson.
   

7. [COLT’19] A new algorithm for non-stationary contextual bandits: Efficient, optimal and parameter-free.
   Yifang Chen, Chung-Wei Lee, Haipeng Luo, Chen-Yu Wei.