Zhengyao Jiang

PhD Student of Machine Learning

University College London

Biography

I’m Zhengyao Jiang, a PhD in Machine Learning student at UCL, supervised by Tim Rocktäschel and Edward Grefenstette. My research focuses on making Reinforcement Learning (RL) work with limited online interaction, which is more reflective of real-world data availability. I work on topics related to this vision, such as offline RL, data-efficient RL, and model-based RL. Before my PhD, I had experience in neural symbolic methods and RL for financial trading.

How to pronounce your name? Zheng = j-uhng, where j as in job; yao = y-aoww, where y as in you; Jiang = gee-ahng. If Zhengyao is still too difficult to pronounce, you can call me Yao.

Interests

Machine Learning
Reinforcement Learning
Deep Learning

Education

Ph.D. Machine Learning, 2020-

University College London
M.Res. Computational Statistics and Machine Learning, 2019--2020

University College London
B.Sc. Artificial Intelligence, 2017--2019

University of Liverpool
B.Sc. Information and Computation Science, 2015--2017

Xi'an Jiaotong-Liverpool University

Featured Publications

Efficient Planning in a Compact Latent Action Space

We propose a novel planning-based sequence modelling method that can scale to high dimensionality state-action space.

Zhengyao Jiang, Tianjun Zhang, Micheal Janner, Yueying Li, Tim Rocktäschel, Edward Grefenstette, Yuandong Tian

Efficient Planning in a Compact Latent Action Space

Graph Backup: Data Efficient Backup Exploiting Markovian Transitions

We propose to treat the transition data of an MDP as a graph, and define a novel backup operator exploiting this graph structure. Comparing to multi-step backup, our graph backup method allows counterfactual credit assignment, and can reduce the variance that comes from stochastic environment dynamics.

Zhengyao Jiang, Tianjun Zhang, Robert Kirk, Tim Rocktäschel, Edward Grefenstette

Graph Backup: Data Efficient Backup Exploiting Markovian Transitions

Publications

Zhengyao Jiang, Tianjun Zhang, Micheal Janner, Yueying Li, Tim Rocktäschel, Edward Grefenstette, Yuandong Tian (2023). Efficient Planning in a Compact Latent Action Space. In ICLR2023.

PDF Cite Code Project

Yicheng Luo, Zhengyao Jiang, Samuel Cohen, Edward Grefenstette, Deisenroth Marc (2023). Optimal Transport for Offline Imitation Learning. In ICLR2023 (spotlight).

PDF Cite

Zhengyao Jiang, Tianjun Zhang, Robert Kirk, Tim Rocktäschel, Edward Grefenstette (2022). Graph Backup: Data Efficient Backup Exploiting Markovian Transitions. In Arxiv.

PDF Cite Code

Zhengyao Jiang, Tianjun Zhang, Robert Kirk, Tim Rocktäschel, Edward Grefenstette (2021). Graph Backup: Data Efficient Backup Exploiting Markovian Data. In DeepRL2021.

PDF Cite

Zhengyao Jiang, Pasquale Minervini, Minqi Jiang, Tim Rocktäschel (2021). Grid-to-Graph: Flexible Spatial Relational Inductive Biases for Reinforcement Learning. In AAMAS2021.

PDF Cite Code

Zhengyao Jiang, Shan Luo (2019). Neural Logic Reinforcement Learning. In ICML2019.

PDF Cite Code

Zhengyao Jiang, Dixing Xu, Jinjun Liang (2017). A Deep Reinforcement Learning Framework for the Financial Portfolio Management Problem. In Arxiv.

PDF Cite Code