Introduction

I am a Departmental Lecturer at the Department of Engineering Science, University of Oxford. I am interested in achieving sample-efficient generalization while maintaining scalable computation. I have a particular interest in learning settings that involve distribution shift, including robustness learning, reinforcement learning, and continual learning.

Email: yangchen Dot pan AT eng DOT ox DOT ac DOT uk

People

Zhiyao Luo (2023-present, PhD, Oxford Univ)
Lam Lam (2023-present, Undergrad, Oxford Univ)
Ziyi Wang (2023-present, Undergrad thesis, Oxford Univ)
Yudong Luo (2022-present, PhD, Univ of Waterloo)
Avery Ma (2021-2023, PhD, Univ of Toronto)
Qingfeng Lan (2019-2022, PhD Univ of Alberta)

To prospective students/collaborators.

If you are currently enrolled as a student at Oxford, feel free to reach out.

Publications

Refereed Publications

* indicates co-first authorship.

• Understanding the robustness difference between SGD and adaptive gradient methods. [paper]
  Avery Ma, Yangchen Pan, Amir-massoud Farahmand.
  Transactions on Machine Learning Research (TMLR, featured certification), 2023.

• An Alternative to Variance: Gini Deviation for Risk-averse Policy Gradient. [paper]
  Yudong Luo, Guiliang Liu, Pascal Poupart, Yangchen Pan.
  Conference on Neural Information Processing Systems (NeurIPS), 2023.

• Memory-efficient Reinforcement Learning with Value-based Knowledge Consolidation. [paper]
  Qingfeng Lan, Yangchen Pan, Jun Luo, A. Rupam Mahmood.
  Transactions on Machine Learning Research (TMLR), 2023.

• Conditionally Optimistic Exploration for Cooperative Deep Multi-Agent Reinforcement Learning. [paper]
  Xutong Zhao, Yangchen Pan, Chenjun Xiao, Sarath Chandar, Janarthanan Rajendran.
  Conference on Uncertainty in Artificial Intelligence (UAI), 2023.

• The In-Sample Softmax for Offline Reinforcement Learning. [paper]
  Chenjun Xiao *, Han Wang *, Yangchen Pan, Adam White, Martha White.
  International Conference on Learning Representations (ICLR), 2023.

• Greedy Actor-Critic: A New Conditional Cross-Entropy Method for Policy Improvement. [paper]
  Samuel Neumann, Sungsu Lim, Ajin George Joseph, Yangchen Pan, Adam White, Martha White.
  International Conference on Learning Representations (ICLR), 2023.

• Understanding and Mitigating the Limitations of Prioritized Experience Replay. [paper]
  Yangchen Pan *, Jincheng Mei *, Amir-massoud Farahmand, Martha White, Hengshuai Yao, Mohsen Rohani, Jun Luo.
  Conference on Uncertainty in Artificial Intelligence (UAI), 2022.

• An Alternate Policy Gradient Estimator for Softmax Policies. [paper]
  Shivam Garg, Samuele Tosatto, Yangchen Pan, Martha White, Rupam Mahmood.
  International Conference on Artificial Intelligence and Statistics (AISTATS), 2022.

• Fuzzy Tiling Activations: A Simple Approach to Learning Sparse Representations Online. [paper]
  Yangchen Pan, Kirby Banman, White Martha.
  International Conference on Learning Representations (ICLR), 2021.

• An implicit function learning approach for parametric modal regression. [paper]
  Yangchen Pan, Ehsan Imani, Martha White, Amir-massoud Farahmand.
  Conference on Neural Information Processing Systems (NeurIPS), 2020.

• Maxmin Q-learning: Controlling the Estimation Bias of Q-learning. [paper]
  Qingfeng Lan, Yangchen Pan, Alona Fyshe, Martha White.
  International Conference on Learning Representations (ICLR), 2020.

• Frequency-based Search-control in Dyna. [paper]
  Yangchen Pan *, Jincheng Mei *, Amir-massoud Farahmand.
  International Conference on Learning Representations (ICLR), 2020.

• Hill Climbing on Value Estimates for Search-control in Dyna. [paper]
  Yangchen Pan, Hengshuai Yao, Amir-massoud Farahmand, Martha White.
  International Joint Conference on Artificial Intelligence (IJCAI), 2019.

• Reinforcement learning with function-valued action spaces for partial differential equation control. [paper]
  Yangchen Pan, Amir-massoud Farahmand, Martha White, Saleh Nabi, Piyush Grover, Daniel Nikovski.
  International Conference on Machine Learning (ICML), 2018.

• Organizing experience: a deeper look at replay mechanisms for sample-based planning in continuous state domains. [paper]
  Yangchen Pan, Muhammad Zaheer, Adam White, Andrew Patterson, Martha White.
  International Joint Conference on Artificial Intelligence (IJCAI), 2018.

• Adapting kernel representations online using submodular maximization. [paper]
  Matthew Schlegel, Yangchen Pan, Jiecao Chen, Martha White.
  International Conference on Machine Learning (ICML), 2017.

• Effective sketching methods for value function approximation. [paper]
  Yangchen Pan, Erfan Sadeqi Azer, Martha White.
  Conference on Uncertainty in Artificial Intelligence (UAI), 2017.

• Accelerated gradient temporal difference learning. [paper]
  Yangchen Pan, Adam White, Martha White.
  AAAI Conference on Artificial Intelligence (AAAI), 2017.

• Incremental truncated LSTD. [paper]
  Clement Gehring, Yangchen Pan, Martha White.
  International Joint Conference on Artificial Intelligence (IJCAI), 2016.

Preprints/Workshop/Work in progress

• Label Alignment Regularization for Distribution Shift. [paper]
  Ehsan Imani, Guojun Zhang, Jun Luo, Pascal Poupart, Philip Torr, Yangchen Pan.

• Sparse NN.

• RL4SL.

PhD thesis

• Improving Sample Efficiency of Online Temporal Difference Learning. Yangchen Pan.

Teaching

2024 Hilary: Machine learning lab, University of Oxford. [website]
2023 Trinity: CWM 10, Artificial Intelligence and Machine Learning with python, University of Oxford [website]
2019 Fall: CMPUT 466/566, Machine Learning, Teaching Assistant, University of Alberta
2019 Spring: CMPUT 272, Formal Systems and Logic in Computing Science, Teaching Assistant, University of Alberta
2016 Spring: CSCI C343, Data Structure, Associate Instructor (aka TA), Indiana University at Bloomington
2015 Fall: CSCI B503, Algorithm Design and Analysis, Associate Instructor (aka TA), Indiana University at Bloomington
2014 Fall: CSCI 1311 Discrete Structure I, Teaching Assistant, George Washington University

Service

Conference Program Committee Member or Reviewer

NeurIPS 2018-present.
ICML 2018-present.
ICLR 2017-present.
AISTATS 2021-present.

Journal Reviewer

Journal of Machine Learning Research (JMLR) 2020, 2021 (co-reviewed), 2022
Transactions on Machine Learning Research (TMLR), 2022-