About Me
I am currently a Ph.D. Candidate at the School of Computing Science at Simon Fraser University, under the supervision of Pro.Oliver Schulte . My research interest lies in 1) Deep Reinforcement Learning and Interpretability, 2) Spatio-Temporal Data Mining, and 3) Variational Models. I have also finished two research internships at Cognitive Computing Lab, Baidu Research (supervised by Pro.Ping Li) and SLiQ Lab, Sportlogiq.
Undergraduate Study: I received my Bachelor degree at South China University of Technology (major in Computing Science). I have spent time pursuing my interest in Mathematics Modeling. Our best score is the National First-class Honor in Contemporary Undergraduate Mathematical Contest in Modeling (CUMCM) The Ministry of Education of the People’s Republic of China gave me a National Scholarship for my academic performance as an undergraduate student. The other worth-mentioning experiences include 1) An research assistant at the Machine Learning and Pattern Recognition Lab, (under the supervision of Dr. Patrick Chan). 2) finished two software engineering internships at Ping An Technology (Shenzhen) and Datatub(Guangzhou).
For my full cv, please visit here
Research
Here list some brief introductions to the research projects that I have worked on.- A Deep Reinforcement Learning Model for Player Evaluation. My first work combined the DRL model with Dynamic Trace length LSTM to model the sequential data in sports. We have built a DRL-based player ranking system and developed a novel metric to estimated the value of players, using the play-by-play Video Tracking dataset provided by sportlogiq.
- Interpret Deep Reinforcement Learning with Mimic Learning. We proposed a Linear Model U-Tree (LMUT) algorithm to distill the knowledge from the opaque deep model to a transparent model with tree structure representation. The rules and knowledge are extracted from LMUT to interpret the knowledge of DRL models.
- Extract knowledge from Web Text with Deep Reinforcement Learning. We improve the performance of Sequence-to-Sequence (Seq2Seq) information extractors with Actor-Critic algorithm and Monte-Carlo Tree Search for the task of Open Information Extraction (OIE).
Selected Publications
- An Advantage Actor-Critic Algorithm with Confidence Exploration for Open Information Extraction.
Guiliang Liu, Xu Li, Mingming Sun, Ping Li. Proceedings SDM 2020.
[Paper (PDF)] - Extracting Knowledge from Web Text with Monte Carlo Tree Search.
Guiliang Liu, Xu Li, Jiakang Wang, Mingming Sun, Ping Li. Proceedings WWW 2020.
[Paper (PDF)], [Slides] - Cracking the Black Box: Distilling Deep Sports Analytics.
Xiangyu Sun, Jack Davis, Oliver Schulte, Guiliang Liu. Proceedings KDD 2020.
[Paper (PDF)], [arXiv], [Slides], [Video], [code] - Deep soccer analytics: Learning an action-value function for evaluating soccer players.
Guiliang Liu, Yudong Luo, Oliver Schulte, and Tarak Kharrat. ECML-PKDD 2020 Journal Track, published at Data Mining and Knowledge Discovery (DMKD).
[Paper (PDF)], [Slides], [Video] - Deep Reinforcement Learning in Ice Hockey for Context-Aware Player Evaluation.
Guiliang Liu, Oliver Schulte. Proceedings IJCAI 2018.
[Paper (PDF)], [arXiv], [Code], [Slides], - Toward Interpretable Deep Reinforcement Learning with Linear Model U-Trees.
Guiliang Liu, Oliver Schulte, Wang Zhu, Proceedings ECML-PKDD 2018.
[Paper (PDF)], [arXiv], [Slides],
Contact
- Email: gla68@sfu.ca
- Address: School of Computing Science, Simon Fraser University, Burnaby, Canada
- Linkedin: Guiliang (Galen) Liu