Avatar

Ziyang Tang

University of Texas at Austin

Biography

I’m a fifth year Ph.D. student at UT Austin in Statistical Learning and AI group, under the advise by Prof. Qiang Liu. My research interest is mainly in approximate inference and reinforcement learning, with a particular focus on designing efficient algorithm for off-policy reinforcement learning. Before joining UT Austin, I got my Master degree of Computer Science in Cornell University and Bachelor degree in Peking University.

Interests

  • Reinforcement Learning
  • Approximate Inference

Education

  • PhD student, 2016 -- current

    University of Texas at Austin

  • Master of Engineering in Computer Science, 2016

    Cornell University

  • Bachelor of Science in Computer Science, 2015

    Peking University

Recent Publications

Ziyang Tang, Yihao Feng, Na Zhang, Jian Peng, Qiang Liu (2020). Off-Policy Interval Estimation with Lipschitz Value Iteration. Advances in Neural Information Processing Systems.

PDF Arxiv Poster Slides

Yihao Feng*, Tongzheng Ren*, Ziyang Tang*, Qiang Liu (2020). Accountable Off-Policy Evaluation With Kernel Bellman Statistics. International Conference on Machine Learning.

PDF Talks Arxiv Slides

Ziyang Tang*, Yihao Feng*, Lihong Li, Dengyong Zhou, Qiang Liu (2019). Doubly Robust Bias Reduction in Infinite Horizon Off-Policy Estimation. International Conference on Learning Representations(ICLR).

PDF Arxiv Talks Poster Slides

Dilin Wang*, Ziyang Tang*, Chandrajit Bajaj, Qiang Liu (2019). Stein Variational Gradient Descent with Matrix-Valued Kernels. Advances in Neural Information Processing Systems.

PDF Poster

Qiang Liu, Lihong Li, Ziyang Tang, Dengyong Zhou (2018). Breaking the curse of horizon: Infinite-horizon off-policy estimation. Advances in Neural Information Processing Systems.

PDF Poster Slides Arxiv

Experience

 
 
 
 
 

Research Intern

Applied Machine Learning(AML) Group at Bytedance

Jun 2020 – Aug 2020 Seattle
Designing a new retrieval algorithm framework that can accelerate the retrieval stage while keeping the accuracy, supervised by Chong Wang