Portfolio item number 1
Short description of portfolio item number 1
Posts by Collection
portfolio
Portfolio item number 2
Short description of portfolio item number 2
publications
On the Linear Convergence of Policy Gradient under Hadamard Parametrization
Published in Information and Inference: A Journal of the IMA, 2023. https://arxiv.org/abs/2305.19575.
Recommended citation: Jiacai Liu, Jinchi Chen, and Ke Wei. On the Linear Convergence of Policy Gradient under Hadamard Parameterization. arXiv:2305.19575, 2023.
On the Convergence of Projected Policy Gradient for Any Constant Step Sizes
Published in arxiv, 2023. https://arxiv.org/abs/2311.01104.
Recommended citation: Jiacai Liu, Wenye Li, and Ke Wei. On the Convergence of Projected Policy Gradient for Any Constant Step Sizes. arXiv:2311.0110, 2023.
Elementary Analysis of Policy Gradient Methods
Published in arxiv, 2024. https://arxiv.org/abs/2404.03372.
Recommended citation: Jiacai Liu, Wenye Li, and Ke Wei. Elementary Analysis of Policy Gradient Methods. arXiv:2404.03372, 2024.
$phi$-Update: A Class of Policy Update Methods with Policy Convergence Guarante
Published in ICLR, 2024. https://openreview.net/pdf?id=fh7GYa7cjO.
Recommended citation: Wenye Li, Jiacai Liu and Ke Wei. $\phi$-Update: A Class of Policy Update Methods with Policy Convergence Guarante. In The Thirteen International Conference on Learning Representations, ICLR 2025, Singapore.
Improving Multi-Step Reasoning Abilities of Large Language Models with Direct Advantage Policy Optimization
Published in arxiv, 2024. https://arxiv.org/pdf/2412.18279.
Recommended citation: Jiacai Liu and Chaojie Wang and Chris Yuhao Liu and Liang Zeng and Rui Yan and Yiwen Sun and Yang Liu and Yahui Zhou. arXiv:2412.18279, 2024.
talks
Some Progress on the Convergence of Policy Gradient Methods
Published:
Abstract: In this talk, I’ll introduce some novel convergence results of the policy gradient method and its variants from my papers. There are three main results I would like to present in this talk: 1. The projected policy gradient converges in a finite number of iterations for any given constant step size. 2. The exact softmax policy gradient converges to the optimum for any given constant step size. 3. $\phi$-update, a class of policy update methods with a policy convergence guarantee that goes beyond the policy gradient descent (or mirror descent) framework.
teaching
Teaching experience 1
Undergraduate course, University 1, Department, 2014
This is a description of a teaching experience. You can use markdown like any other post.
Teaching experience 2
Workshop, University 1, Department, 2015
This is a description of a teaching experience. You can use markdown like any other post.