Yaohua Liu (刘耀华)

Yaohua Liu (刘耀华)

Ph.D.

Biography

Introduction

I am currently a Postdoctoral Fellow at the Visual AI Lab, The University of Hong Kong (HKU), working with Prof. Kai Han.

I received my Ph.D. degree in Software Engineering from Dalian University of Technology (DLUT) in June 2025, under the supervision of Prof. Risheng Liu (National Excellent Young Scholar). Prior to that, I conducted research at the Key Laboratory for Ubiquitous Network and Service Software of Liaoning Province.

My research focuses on Optimization for Machine Learning, with a particular emphasis on Bilevel Optimization (BLO) and its applications in Trustworthy AI. I am dedicated to developing reliable and efficient algorithms to address complex challenges in hierarchical learning, security, and multi-modal systems.

Currently, my research interests center on the following areas:

  • Optimization for Machine Learning: Methodology and analysis of Bilevel Optimization, Hyperparameter Optimization, and Neural Architecture Search (NAS).
  • Trustworthy Machine Learning & Security: Adversarial Robustness, Black-box Transfer Attacks, and safety alignment in Large Multi-modal Models (LMMs).
  • Intelligent Agents: Optimization frameworks for Mobile/GUI Agents and their security implications.

My research outcomes have appeared in representative journals and conferences, including IEEE TPAMI, IEEE TNNLS, NeurIPS, ICML, CVPR, AAAI, IJCAI, and ACM MM. I also serve as a reviewer for venues such as CVPR, ECCV, AAAI, and IJCAI.

Recent News

  • [12/2025] New! Our paper “Augmenting Iterative Trajectory for Bilevel Optimization: Methodology, Analysis and Extensions” has been accepted by IEEE TPAMI.
  • [11/2024] One paper on Tri-level Learning Framework for Image Enhancement was accepted to ACM MM 2024.
  • [10/2024] One paper on Dual-Stream-Modulated Learning was accepted to IEEE TNNLS.
  • [09/2024] One paper titled “Collaborative Brightening and Amplification of Low-Light Imagery Via Bi-Level Adversarial Learning” was accepted to Pattern Recognition (PR).
  • [08/2024] One paper on Generalized Transfer Attack with Bilevel Optimization was accepted to IJCAI 2024.
  • [05/2023] One paper titled “Averaged method of multipliers for bi-level optimization without lower-level strong convexity” was accepted to ICML 2023.
  • [08/2023] One paper titled “PEARL: Preprocessing Enhanced Adversarial Robust Learning of Image Deraining for Semantic Segmentation” was accepted to ACM MM 2023.

Contact

I am always open to academic discussions and potential collaborations, particularly regarding Optimization for Machine Learning, Trustworthy AI Security, and Multi-modal Models.