Yuheng Bu Assistant Professor Department of Computer Science (CS) College of Engineering Univ. of California - Santa Barbara Contact Information Email: buyuheng (at) ucsb (dot) edu Google Scholar DBLP

Biography

I am an Assistant Professor in the Department of Computer Science (CS) at UC Santa Barbara.

Before joining the UC Santa Barbara, I was an assistant professor in the Department of Electrical and Computer Engineering (ECE) at the University of Florida. Prior to that, I was a postdoctoral research associate at the Research Laboratory of Electronics (RLE) and Institute for Data, Systems, and Society (IDSS), Massachusetts Institute of Technology (MIT). I received my Ph.D. in 2019 from the Coordinated Science Laboratory and the Department of Electrical and Computer Engineering, University of Illinois at Urbana-Champaign (UIUC). Earlier, I earned my B.S. degree (with honors) in Department of Electronic Engineering from Tsinghua University in 2014.

Research

My research interests lie in the intersection of machine learning, information theory and signal processing. I leverage tools from information theory and signal processing to develop theoretically justified learning algorithms for applications including watermarking for generative AI, fair machine learning, uncertainty quantification, and model compression.

More broadly, the primary goal of my research is to lay information-theoretic foundations for trustworthy learning algorithms, particularly with generalization, fairness, and robustness guarantees.

To Prospective Students

For all students interested in our group: please fill out this Google Form. I will carefully review submissions, though I may not be able to respond to every email individually.

PhD candidates: I am recruiting students for Fall 2026. I seek self-motivated students with strong backgrounds in machine learning, statistics, and information theory. You should be comfortable reading and writing proofs and learning challenging new mathematical concepts; in short, mathematical maturity is essential.

Selected Publications

• Y. Liu, Y. Song, H. Ci, Y. Zhang, H. Wang, Z. Shou, Y. Bu. “Image Watermarks are Removable using Controllable Regeneration from Clean Noise,” International Conference on Learning Representations (ICLR), Apr. 2025.

• Y. Liu, Y. Bu. “Adaptive Text Watermark for Large Language Models,” International Conference on Machine Learning (ICML), Jul. 2024.

• M. Shen, Y. Bu, G. W. Wornell. “On Balancing Bias and Variance in Unsupervised Multi-Source-Free Domain Adaptation,” International Conference on Machine Learning (ICML), Jul. 2023.

• G. Aminian*, Y. Bu*, L. Toni, M. R. D. Rodrigues, G. W. Wornell. “An Exact Characterization of the Generalization Error for the Gibbs Algorithm,” (* equal contribution), Conference on Neural Information Processing Systems (NeurIPS), 2021.

• Y. Bu*, J. K. Lee*, D. Rajan, P. Sattigeri, R. Panda, S. Das, G. W. Wornell. “Fair Selective Classification via Sufficiency,” (* equal contribution), International Conference on Machine Learning (ICML), Jul. 2021. (Long Talk, Top 3%)

• Y. Bu, S. Zou, V. V. Veeravalli. “Tightening Mutual Information Based Bounds on Generalization Error,” IEEE Journal on Selected Areas in Information Theory, vol. 1, pp. 121 - 130, May 2020.

News

• Jul. 2025, our paper, “Defending LLM Watermarking Against Spoofing Attacks with Contrastive Representation Learning,” has been accepted by COLM 2025 (acceptance rate: 32%)!

• Jun. 2025, our paper, “Class-wise Generalization Error: An Information-Theoretic Analysis,” has been accepted by Transactions on Machine Learning Research!

• May 2025, our paper, “Fairness Overfitting in Machine Learning: An Information-Theoretic Perspective,” has been accepted by ICML 2025 (acceptance rate: 27%)!

• Mar. 2025, I have co-organized the 8th Annual Workshop on Cognition & Control together with Prof. Sean Meyn and Prof. Jie Fu at UF Reitz Union.

• Jan. 2025, our paper, “Image Watermarks are Removable using Controllable Regeneration from Clean Noise,” has been accepted by ICLR 2025 (acceptance rate: 32%)!

• Nov. 24, 2024, I will present the tutorial “Characterizing the Generalization Error of Machine Learning Algorithms Via Information Measures” at the IEEE Information Theory Workshop (ITW 2024), Shenzhen, China! Jointly with Gholamali Aminian (The Alan Turing Institute, UK), Samir M. Perlaza (INRIA, France), and Iñaki Esnaola (University of Sheffield, UK).
  All Slides, Slides Part I (Yuheng Bu), Slides Part II (Yuheng Bu), Slides Part III (Samir Perlaza), Slides Part IV (Samir Perlaza)

• Sep. 2024, our paper, “Are Uncertainty Quantification Capabilities of Evidential Deep Learning a Mirage?” has been accepted by NeurIPS 2024 (acceptance rate: 26%)!

• Jul. 2024, our paper, “Information-theoretic Analysis of the Gibbs Algorithm: An Individual Sample Approach,” has been accepted by ITW 2024. Additionally, we will organize a tutorial at ITW 2024 titled “Characterizing the Generalization Error of Machine Learning Algorithms Via Information Measures.” Stay tuned!

• Jun. 2024, our paper, “Learning Orthonormal Features in Self-Supervised Learning using Functional Maximal Correlation,” has been accepted by ICIP 2024.

• May 2024, two papers accepted at ICML 2024 (acceptance rate: 28%)!

  • The first paper, “Adaptive Text Watermark for Large Language Models”, proposes a watermarking technique that distinguishes AI/human-generated text. This is Yepeng's first paper since he joined the group! Congrats!

  • The second paper, “Operator SVD with Neural Networks via Nested Low-Rank Approximation”, proposes a new optimization framework that can train neural networks to learn the top singular functions of a given operator.

• Apr. 2024, our paper, “Information-Theoretic Opacity-Enforcement in Markov Decision Processes”, has been accepted by IJCAI 2024.

• Apr. 2024, two papers, “Towards Optimal Inverse Temperature in the Gibbs Algorithm” and “Group Fairness with Uncertain Sensitive Attributes,” have been accepted by ISIT 2024.

• Mar. 2024, I presented our paper “Adaptive Text Watermark for Large Language Models” at Annual Conference on Information Science and Systems (CISS), Princeton, NJ.

• Jan. 2024, I have co-organized the 7th Annual Workshop on Cognition & Control together with Prof. Sean Meyn and Prof. Jie Fu at UF Reitz Union. Talks include topics such as reinforcement learning, accelerated optimization, information theory, brain-machine interface systems, and much more!

• Jan. 2024, our paper, “Gibbs-Based Information Criteria and the Over-Parameterized Regime”, has been accepted by AISTATS 2024 (acceptance rate: 28%)! This is my student Haobo's first paper, Congrats to Haobo!