The CSE Best PhD Dissertation Award for 2024/25

We are pleased to announce that the recipient of the CSE Best PhD Dissertation Award is Dr. Pingchuan Ma for his work titled "Algorithms, Applications, and Verification of Causal Structure Learning." The dissertation was supervised by Prof. Shuai Wang.

In addition, we would like to recognize the following students for their outstanding dissertations, which were selected as Honourable Mentions:

1. Dr. Haoqiang HUANG for his work on "Exact and Approximate Algorithms for Frechet Problems," advised by Prof. Siu-Wing CHENG.
2. Dr. Yi LIN for his work on "Toward Efficient Yet Effective Algorithms for Medical Image Segmentation," advised by Prof. Hao CHEN.
3. Dr. Kaiqiang XU for his work on "Towards Efficient and Accessible Systems for Distributed Machine Learning," advised by Prof. Kai CHEN.

We congratulate all of our awardees and thank everyone who submitted dissertations for consideration. We are proud of our students' excellent performance and appreciate their contributions to our department's academic community.

About Dr. Pingchuan Ma's Dissertation and His Sharing

Biography

Dr. Pingchuan MA is a Professor in the College of Computer Science and Technology at Zhejiang University of Technology. His research lies at the intersection of data systems, causal inference, and privacy-enhancing computation. He is also the co-founder of CipherInsight Limited, a HKUST spin-off focusing on verifiable and privacy-preserving analytics. His research has been recognized by multiple international venues such as SIGMOD, ICSE, KDD, and VLDB, and he has received honors including the HKUST SENG PhD Research Excellence Award and the Hong Kong ICT Bronze Award.

Dr. Pingchuan MA

Research Overview

Dr. MA's dissertation, Algorithms, Applications, and Verification of Causal Structure Learning, develops a principled framework for discovering causal relationships from complex data and ensuring their correctness through formal verification. It bridges causal inference, software engineering, and data systems to build trustworthy, explainable, and efficient causal discovery pipelines.

Key Breakthroughs

The dissertation presents three connected lines of work that collectively enhance the reliability and scalability of causal discovery.

• It introduces ML4S (KDD 2022) and SPOT (KDD 2024), two methods leveraging machine learning and amortized inference to recover large causal graphs efficiently and robustly.
• It proposes XInsight (SIGMOD 2023) and Guardrail (SIGMOD 2026), which integrate causal reasoning with program synthesis for explainable and integrity-aware data analysis.
• It develops CICheck (ICSE 2024) and its extensions ED-Check and P-Check, transforming conditional independence reasoning into automated runtime verification for error detection and privacy assurance.

Together, these contributions form a coherent foundation for building scalable, verifiable, and trustworthy causal reasoning systems.

Recent Work

Recently, Dr. MA and his collaborators have extended this line of research toward privacy-preserving causal reasoning, combining causal inference with zero-knowledge proofs to ensure analytical integrity without exposing sensitive data. This work, in collaboration with industry partners, aims to enable verifiable data intelligence for financial and healthcare applications and has been recognized for the Hong Kong ICT Bronze Award.

Acknowledgements

"HKUST CSE has created a transparent and open academic environment that allows students to grow and fully develop their potential. The department encourages collaboration across research groups and disciplines, and its supportive culture enables students to explore ideas with confidence. I am especially grateful for the guidance of my advisor, Prof. Shuai WANG. Prof. WANG has encouraged me whenever I encountered challenges and has provided valuable opportunities for me to mentor junior students, which helped me mature both academically and personally. Prof. WANG is, in my view, the best supervisor, educator and role model I could ever hope for. I believes it will take me decades to learn from Prof. WANG's professionalism, intellectual rigor and admirable personality. I would also like to express sincere gratitude to all collaborators whose support and cooperation have enriched my research journey."

Personal Reflection

"I am deeply honoured to receive the CSE Best PhD Dissertation Award. This recognition carries profound meaning for me because it reflects not only years of dedicated research but also the continuous support, trust and encouragement I has received from my advisor, collaborators and the department. The award reminds me of the countless discussions, experiments and revisions that shaped my work, and it highlights the value of perseverance in pursuing impactful research. It also strengthens my determination to continue contributing to the scientific community. I hopes to advance responsible and trustworthy data technologies that can create long-term positive influence, both in academic research and in real-world applications. This award serves as an inspiration for me to uphold higher standards in integrity, curiosity and collaboration throughout my future career."

About Haoqiang HUANG's Dissertation and His Sharing

Biography

Dr. Haoqiang HUANG is currently a researcher with the Huawei TopMinds program at the Huawei Beijing Research Centre. He earned his PhD in 2025 from the Department of Computer Science and Engineering at the Hong Kong University of Science and Technology (HKUST), where Dr. HUANG was supervised by Prof. Siu-Wing Cheng. Prior to that, he completed his Bachelor's degree in 2019 in the Department of Computer Science and Technology at the University of Science and Technology.

Dr. Haoqiang HUANG

Research Overview

Dr. Huang's dissertation provides a systematic study of the Fréchet distance in trajectory data analysis. While widely used as a trajectory similarity measure, theoretical understanding lags in computational methods, data management, and real-world adaptations. His work advances these areas.

Key Breakthroughs

Beyond developing new algorithms for efficient and effective Fréchet distance use in trajectory analytics, Dr. Huang designed the fastest known algorithm for its computation to date. This marks the first major breakthrough since 1995.

Acknowledgements

"My supervisor played an instrumental role in my growth as a researcher. Through our discussions, I acquired valuable problem-solving skills and research techniques. He set an exemplary model for rigorous and impactful research. I am also grateful for the department's collaborative and supportive environment, which encouraged open communication between professors and peers. Regular seminars enriched my perspective by exposing me to diverse research fields."

Personal Reflection

"I am truly honoured to receive this award. It represents significant recognition of my five years of research efforts and the most meaningful gift I could ask for. I will cherish this honour and use it as motivation to produce high-quality research."

About Yi LIN's Dissertation and His Sharing

Biography

Dr. Yi LIN is currently a postdoctoral associate at the Department of Population Health Sciences, Weill Cornell Medicine. He obtained his PhD degree from the Department of Computer Science and Engineering (CSE) at The Hong Kong University of Science and Technology (HKUST), where he was supervised by Prof. Hao CHEN and Prof. Kwang-Ting CHENG. Prior to his PhD, Dr. LIN worked as a research scientist in Jarvis Lab at Tencent. He received his M.E. and B.E. degrees from Huazhong University of Science and Technology (HUST) in 2020 and 2016, respectively. His research interests include medical image analysis, artificial intelligence, and multi-modal learning. His work has been published in top-tier conferences and journals such as TMI, MedIA, TCYB, MICCAI, and IPMI. His PhD dissertation was awarded the first place in the Edward H. Shortliffe Doctoral Dissertation Award 2025. He serves as a reviewer for more than 30 top-tier journals and conferences including TPAMI, TCSVT, TNNLS, TMI, MedIA, IF, JBHI, CVPR, ECCV, ICCV, and MICCAI. His expertise in medical image analysis is further demonstrated by championship awards in three international competitions: the MICCAI AASCE Challenge, the MICCAI CADA Challenge, and the Tencent Medical AI Challenge.

Dr. Yi LIN

Research Overview

Dr. Lin's thesis explores three interrelated topics in effective and efficient medical image analysis: data-efficient learning, label-efficient learning, and model-efficient learning. These address key challenges in developing deep learning models for clinical applications.

Key Breakthroughs

The dissertation's contributions unfold in three areas:

1. Advanced data-efficient learning for medical image segmentation, introducing methods that promote data diversity and substantially reduce training samples.
2. Enhanced label efficiency via a novel weakly supervised approach, cutting annotation needs to just 1% of fully supervised methods while boosting practicality.
3. Model-efficient learning through neural architecture optimization and knowledge distillation, lowering computational costs while maintaining performance for resource-limited settings.

Recent Work

Post-PhD, Dr. Lin has focused on novel deep learning methods for multi-modal medical image analysis, including:

• Vision-language models for medical image report generation;
• Multi-modal data synthesis and augmentation;
• Self-supervised learning for multi-modal medical images.

Acknowledgements

"First and foremost, I wish to express my deepest gratitude to my supervisor, Prof. Hao CHEN, whose continuous support, guidance, and encouragement have been instrumental throughout my doctoral studies. I remain deeply appreciative of his patience, motivation, and enthusiasm, as well as his exceptional breadth of knowledge. His discerning advice, perceptive feedback, and constructive critiques have fundamentally shaped and elevated the quality of my research. His mentorship has been invaluable to both the progression of my investigations and the composition of this thesis.

I am equally indebted to my co-supervisor, Prof. Kwang-Ting CHENG, for his invaluable counsel, insightful recommendations, and thoughtful criticism. His steadfast encouragement and support have profoundly impacted my academic journey, and I shall always cherish his integrity and meticulousness, which are the goals I strive for throughout my life.

I am grateful to the Department of Computer Science and Engineering at HKUST for cultivating a dynamic research environment, affording access to cutting-edge facilities, and fostering collaborations with distinguished experts in the field. I extend my sincere thanks to the members of my thesis committee, Prof. Jin QIN, Prof. Dan XU, Prof. Long CHEN, and Prof. Terence Tsz Wai WONG, for their thoughtful feedback and constructive suggestions that helped improve my research work."

Personal Reflection

"Receiving Honourable Mentions for the CSE Best PhD Dissertation Award is a tremendous honour and a significant milestone in my academic career. I am deeply grateful for the recognition of my research efforts and the opportunity to contribute to the field of medical image analysis. This award not only validates the hard work and dedication I have invested in my dissertation but also inspires me to continue pursuing excellence in my future research endeavours."

About Kaiqiang XU's Dissertation and His Sharing

Biography

Dr. Kaiqiang XU received his PhD in Computer Science and Engineering at HKUST, supervised by Prof. Kai Chen. His research focuses on computer systems for artificial intelligence, including GPU cluster management, carbon-aware cluster scheduling, and privacy-preserving machine learning, with first-author publications at NSDI, ASPLOS, SIGMOD and other top-tier venues. He also led the design and operation of HKUST's Turing AI Computing Cloud (TACC), a shared GPU platform serving over six hundreds of students and researchers across the university.

Dr. Kaiqiang XU

Research Overview

Dr. Xu's dissertation examines building efficient, accessible, and sustainable AI infrastructure. As AI models grow larger, they demand massive GPU clusters, yet current systems waste resources, overlook carbon impact, and hinder secure collaboration. It proposes designs for shared GPU clusters, carbon-aware scheduling, privacy-preserving machine learning, and large-scale graph learning to make AI computing fairer, easier to use, and environmentally friendlier.

Key Breakthroughs

The dissertation introduces four core components:

1. A GPU cluster management system (ASPLOS '25) that improves fairness, utilization, and usability in shared clusters, supporting over 600 users and nearly 30,000 AI jobs at HKUST.
2. The GREEN scheduler (NSDI '25), reducing data-center carbon emissions for AI training by over 40% by shifting workloads to greener energy periods without performance loss.
3. The Sequoia framework (SIGMOD '25), cutting programming effort for privacy-preserving machine learning by over 90% with multi-fold performance gains, enabling secure collaboration in data-sensitive domains.
4. A system for large-scale graph neural networks (SIGMOD '23) that trains billion-edge graphs more efficiently than prior work.

These push AI systems toward faster, fairer, greener, and more secure infrastructure.

Recent Work

Dr. Xu has extended this research across AI infrastructure layers, leading to first-author publications at NSDI, ASPLOS, and SIGMOD, plus contributions to cross-GPU communication spanning distributed systems, architecture, and data management. His GPU cluster design powers HKUST's TACC, supporting university-wide AI teaching and research as a real-world testbed.

Acknowledgements

"The CSE Department and my supervisor have been instrumental throughout this journey. The department provided an open, collaborative environment for learning across systems, architecture, and data management. Prof. Kai Chen offered high-level vision, freedom to explore ambitious ideas, and involvement in real systems like the TACC GPU cluster, grounding research in practical needs. The department's investment in shared infrastructure and cross-group collaboration enabled work spanning subfields and transitioning from papers to practice."

Personal Reflection

"I feel very grateful and humbled to receive this Honourable Mention. It encourages me personally but reflects the collective efforts of my supervisor, collaborators, and the HKUST CSE community. For a systems researcher, seeing ideas deployed in real clusters and recognized by the department is especially meaningful. This award motivates me to develop AI infrastructure benefiting more users and communities."