Welcome to my homepage! I'm Chong Liu, a Ph.D. candidate in computer science advised by Prof. Yu-Xiang Wang at University of California, Santa Barbara. I'm an editorial board reviewer of the Journal of Machine Learning Research (JMLR). I'm also an FAA-certified private pilot with airplane single engine land rating (PP-ASEL). "Chong Liu" is pronounced approximately as "ch-aw-ng l-yo". chongliu [at] cs.ucsb.edu I'm on the job market for Fall 2023!

[Publications] [Impact] [Experiences] [Miscellaneous]

Research Interests

My research focuses on global optimization, decision making, and active learning, as well as their applications in deep neural network hyper-parameter tuning and material discovery. My past research includes ensemble learning, crowdsourcing, and multi-label learning.

News

• Oct. 2022: Paper "Human-in-the-Loop Video Semantic Segmentation Auto-Annotation" is accepted by the 10th IEEE/CVF Winter Conference on Applications of Computer Vision (WACV-2023)! In this paper, we apply active learning to build an auto-annotation framework for video semantic segmentation. Congratulations to my collaborators at Amazon!

• Oct. 2022: I will attend NeurIPS-2022. Happy to talk to you in person in New Orleans, LA!

• Sep. 2022: Happy to be invited to review for AAAI-2023 and AISTATS-2023!

• Jul. 2022: I am attending ICML-2022. Happy to talk to you in person in Baltimore, MD!

• Jun. 2022: Two papers on global optimization and experimental design are accepted by the ICML Workshop on Adaptive Experimental Design and Active Learning in the Real World (ReALML-2022)!

• Apr. 2022: Happy to be invited to speak at RPI CS Theory Seminar!

• Jan. 2022: Paper "Doubly Robust Crowdsourcing" has been published in the Journal of Artificial Intelligence Research (JAIR)! In this paper, crowdsourcing test-time label aggregation is formulated as a statistical estimation problem of inferring the expected voting score. Doubly robust estimator is used to improve crowdsourcing accuracy and save labeling cost at the same time.

• Nov. 2021: Paper "Revisiting Model-Agnostic Private Learning: Faster Rates and Active Learning" has been published in the Journal of Machine Learning Research (JMLR)! In this paper, active learning-based algorithm PATE-ASQ is able work in practice while saving privacy budget at the same time under the differential privacy framework.

Contact

Chong Liu
2118 Henley Hall
University of California, Santa Barbara
Santa Barbara, CA 93106