Mar 2024: Paper on Indexed Minimum Empirical Divergence-Based Algorithms for Linear Bandits the Transactions on Machine Learning Research. Joint work with Ph.D. student Jie Bian. In this paper, we extend the IMED algorithm so that it is amenable to stochastic linear bandits. We show a state-of-the-art regret bound. Congrats Jie Bian!
Mar 2024: Paper on Adversarial Combinatorial Bandits with Switching Costs accepted to the IEEE Transactions on Information Theory. Joint work with former postdoc Yanyan Dong. In this paper, we consider the adversarial combinational bandits problem where the switch of each selected arm in each round incurs a fixed cost. We quantify the regret in the semi-bandit and bandit feedback settings. Congrats Yanyan!
Feb 2024: Paper on DragDiffusion: Harnessing Diffusion Models for Interactive Point-based Image Editing accepted to CVPR 2024 in Seattle and to be presented as a spotlight poster! Congrats to Yujun Shi and all co-authors. In this paper, we enable "drag" editing on diffusion models. By leveraging large-scale pre-trained diffusion models, we greatly improve the generality of "drag" editing.
Jan 2024: Received the Faculty of Science "Faculty Award for Mentorship Excellence (Research Category)".
Jan 2024: Posted a paper on Fixed-Budget Differentially Private Best Arm Identification. This is joint work with ISEM Ph.D. student Zhirui Chen, postdoc P. N. Karthik and ISEM collaborator Yeow Meng Chee. In this paper, we address the problem of best arm identification in the fixed-budget setting under differential privacy constraints, considering the notion of ε-differential privacy and provide guarantees on the error probability. This paper will be presented at the upcoming ICLR in Vienna.
Jan 2024: Three papers accepted at the 2024 International Conference on Learning Representations (ICLR). See CS Conference Papers for details.
Jan 2024: Former Ph.D. student Zhaoqiang Liu and former postdoc Yonglong Li joined UESTC and Xi'an Jiaotong University, respectively, both as 优秀青年科学基金项目(海外)professors.
Nov 2023: Paper on Federated Best Arm Identification with Heterogeneous Clients accepted to the IEEE Transactions on Information Theory. Congrats to Ph.D. student Zhirui Chen and postdoc P. N. Karthik. In this paper, we show, among other cool results, that for any algorithm whose upper bound on the expected time to find the best arms matches with the lower bound up to a multiplicative constant, the ratio of any two consecutive communication time instants must be bounded.
Online Decision Making, Multi-Armed Bandits, Reinforcement Learning Information Theory with Applications to Machine Learning Statistical Signal Processing
There are also multiple positions for talented
postdoctoral scholars. Postdoctoral scholars with strong publication records
and showing interest in the above research topics are also encouraged to
contact me to check with me if there are available positions. Please see this advertisement as well as this.