I am a third year Ph.D. Candidate in Department of Statistics, The Pennsylvania State University. My research lies at the intersection of Reinforcement Learning (RL) and Federated Learning (FL), with a dual focus on:

1. Theoretical Foundations: Designing provable and computationally efficient algorithms for single-agent and federated RL with applications to healthcare and autonomous driving.

2. Practical Applications: Developing models for complex natural systems, including EEG-based neural decoding for biomedical applications and AI-driven heat-alert systems for climate resilience.

You can find my CV here.

I am very furtunate to be advised by Prof. Lingzhou Xue from Department of Statistics, The Pennsylvania State University.

🔥 News

• 2026.1:   A paper is accepted by ICLR 2026.
• 2025.10:   A paper is accepted by Neurips 2025.
• 2025.05:   A paper is accepted by ICML 2025.
• 2025.04:   I attended ICLR 2025 in Singapore.
• 2024.11:   Two papers are accepted by ICLR 2025.

📝 Publications

(* denotes euqal contribution)

1. Q-Learning with Fine-Grained Gap-Dependent Regret.
   Haochen Zhang, Zhong Zheng, and Lingzhou Xue. (2026)
   The Fourteenth International Conference on Learning Representations (ICLR)
   Available at OpenReview and arXiv.

2. Regret-Optimal Q-Learning with Low Cost for Single-Agent and Federated Reinforcement Learning.
   Haochen Zhang*, Zhong Zheng*, and Lingzhou Xue. (2025)
   The Thirty-Ninth Annual Conference on Neural Information Processing Systems (NeurIPS).
   Available at OpenReview and arXiv.

3. Gap-Dependent Bounds for Federated Q-Learning.
   Haochen Zhang*, Zhong Zheng*, and Lingzhou Xue (2025).
   The Forty-second International Conference on Machine Learning (ICML).
   Available at OpenReview and arXiv.

4. Gap-Dependent Bounds for Q-Learning using Reference-Advantage Decomposition.
   Zhong Zheng*, Haochen Zhang*, and Lingzhou Xue (2025).
   The Thirteenth International Conference on Learning Representations (ICLR).
   (Spotlight, 3.26% acceptance rate)
   Available at Openreview and arXiv.

5. Federated Q-Learning with Reference-Advantage Decomposition: Almost Optimal Regret and Logarithmic Communication Cost.
   Zhong Zheng*, Haochen Zhang*, and Lingzhou Xue (2025).
   The Thirteenth International Conference on Learning Representations (ICLR).
   Available at Openreview and arXiv.

🎖 Honors and Awards

• Gold Medal, 2018 Chinese Mathematical Olympiad (CMO)
• Second Prize, 2020 National Undergraduate Mathematics Competition of China (Class A)
• Second Prize, 2021 National Undergraduate Mathematics Competition of China (Class A)

📖 Educations

Ph.D. in Statistics
The Pennsylvania State University, 2023–Present
Advisor: Dr. Lingzhou Xue

M.Sc. in Statistics
The Pennsylvania State University, 2023–2025
Advisor: Dr. Lingzhou Xue
Thesis: Gap-Dependent Regret for Federated Q-Learning

B.Sc. in Statistics
Peking University, 2019–2023