刘勇 Yong Liu (Y. Liu) Ph.D. Candidate HPC-AI LAB School of Computing, Department of Computer Science National University of Singapore, Singapore. Email: lucasliunju [at] gmail.com

Latest News

• [2022.09] One paper is accepted by NeurIPS 2022.

• [2022.03] Two papers are accepted by ICDE 2022.

• [2022.03] One paper is accepted by CVPR 2022.

• One paper is accepted by ICLR 2022.

• One paper is accepted by AAAI 2022.

Short Bio

• From January 2021, Yong will start his Ph.D. degree in Machine Learning and High Performance Computing under the supervision of Prof. Yang You

• Yong received his B.Sc. degree from China Agricultural University, China in June 2017. After that, he became an M.Sc. student in the R&L Group led by Prof. Yang Gao in Nanjing University.

Main Research Interests

Large-Batch Training

• Yong focus on Large-Batch Training on large-scale distributed systems.
  We can use Large-Batch Training to accelerate the training process of deep neural networks.

Multi-Agent Systems

• Yong focus on simplification of learning process in multi-agent systems, i.e., game abstraction.
  The large number of agents and complex game relationship cause great difficulty for policy learning. Therefore, simplifying the learning process is an important research issue.

Reinforcement Learning

• Yong focus on the algorithm framework of reinforcement learning and applications, especially in multi-agent systems. Reinforcement learning learn the policy through the feedback from environment, which is a way closer to the human model. I'm really interested in it.

Transfer Learning

• Yong focus on the transfer learning in multi-agent systems, especially between environments with different number of agents. Policy learning is difficult in large-scale multi-agent systems. We use incremental learning and teansfer learning to solve the problem.

Publications

Yong Liu, Siqi Mai, Minhao Cheng, Xiangning Chen, Cho-Jui Hsieh and Yang You. Random Sharpness-Aware Minimization. NeurIPS'2022. (To appear)

Yong Liu, Siqi Mai, Xiangning Chen, Cho-Jui Hsieh and Yang You. Towards Efficient and Scalable Sharpness-Aware Minimization. CVPR'2022. [Paper]

Yong Liu, Xiangning Chen, Minhao Cheng, Cho-Jui Hsieh and Yang You. Concurrent Adversarial Learning for Large-Batch Training. ICLR'2022. [Paper]

Fuzhao Xue, Ziji Shi, Futao Wei, Yuxuan Lou, Yong Liu, Yang You. Go Wider Instead of Deeper. AAAI'2022. [Paper]

Yong Liu, Xinghua Zhu, Jianzong Wang, Jing Xiao. A Quantitative Metric For Privacy Leakage in Federated Learning. ICASSP'2021, Toronto, Canada, 2021. [Paper]

Yong Liu, Yujing Hu, Yang Gao, Yingfeng Chen, Changjie Fan. Value Function Transfer for Deep Multi-Agent Reinforcement Learning Based on N-Step Returns. In: Proceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI'19), Macau, China, 2019. [Paper] A novel MDP similarity measure and transfer single-agent knowledge to multi-agent environments.

Yong Liu, Weixun Wang, Yujing Hu, Jianye Hao, Xingguo Chen, Yang Gao. Multi-Agent Game Abstraction via Graph Attention Neural Network. In: Proceedings of the 34th AAAI Conference on Artificial Intelligence (AAAI'20), New York, USA, 2020. [Paper] A novel Game Abstraction based on two-stage attention and graph neural network.

Weixun wang, Tianpei Yang, Yong Liu, Jianye Hao, Xiaotian Hao, Yujing Hu, Yingfeng Chen, Changjie Fan, Yang Gao. Action Semantics Network: Considering the Effects of Actions in Multiagent Systems. In: International Conference on Learning Representations (ICLR'20), Addis Ababa, Ethiopia, 2020. [Paper] We propose a novel network architecture, named Action Semantics Network (ASN), that explicitly representssuch action semantics between agents. ASN characterizes different actions' influence on other agents using neural networks based on the action semantics between agents.

Weixun wang, Tianpei Yang, Yong Liu, Jianye Hao, Xiaotian Hao, Yujing Hu, Yingfeng Chen, Changjie Fan, Yang Gao. From Few to More: Large-scale Dynamic Multiagent Curriculum Learning. In: Proceedings of the 34th AAAI Conference on Artificial Intelligence (AAAI'20), New York, USA, 2020. [Paper] We propose three transfer mechanisms across curricula to accelerate the learning process. Moreover, due to the fact that the state dimension varies across curricula, and existing network structures cannot be applied in such a transfer setting. Therefore, we design a novel network structure called Dynamic Agent-number Network (DyAN).

Shaokang Dong, Jiarui Chen, Yong Liu, Tianyi Bao, Yang Gao. Reinforcement Learning from Algorithm Model to Industry Innovation: A Foundation Stone of Future Artificial Intelligence. ZTE Communications (accepted) We introduce the current research progress and application of reinforcement learning.

• Yinghuan Shi, Tiexin Qin, Yong Liu, Jiwen Lu, Yang Gao. Data Augmentation for Small-Scale Data by Learning the Deterministic Policy. arXiv:1910.08343.

• Zemian Ke, Pan Liu, Yong Liu, Zhibin Li. A Double Deep Q Network-based Variable Speed Limit Control to Reduce Travel Time at Freeway Bottlenecks. submitted to Transportation Research Record.

Awards & Honors

• JSAI Outstanding Master's Thesis Award Honorable Mentaion (top 10 in JSAI) [link]

• Excellent Master Thesis in NJU CS, 2020

• National Scholarship, 2019

• Beijing Excellent Graduates, 2017

• National Scholarship, 2016

• Interdisciplinary Contest In Modeling (ICM), Honorable Mention, 2016

• Beijing University Physics Experiment Competition, Third Award, 2014

Correspondence

Email: liuyong [at] comp.nus.edu.sg
Office: Room i4-02-07, National University of Singapore
Address: Yong Liu
                 National Key Laboratory for Novel Software Technology
                 Nanjing University, Xianlin Campus Mailbox 603
                 163 Xianlin Avenue, Qixia District, Nanjing 210046, China

Last modified by Yong Liu on March. 2, 2022.