AI School
| Time | Sessions | Lecturers |
|---|---|---|
| 08:00 - 17:00 | Registration | |
| 08:25 – 08:30 | Introduction to AI School | A/Prof Guanfeng Liu |
| 08:30 – 09:30 | Lecture 1: Diffusion models: foundations and innovations | Prof James Kwok |
| 09:30 – 10:30 | Lecture 2: Temporal Learning in Healthcare - How to Care of the Missing Data | Prof Myra Spiliopoulou |
| 10:30 – 11:00 | Tea Break (Doric Room) | |
| 11:00 – 12:00 | Lecture 3: | Prof Vincent S. Tseng |
| 12:00 – 12:30 | Lecture 4: Retrieval Augmented Generation (RAG) and Beyond: Boosting the Performance of LLMs | Prof Jae-Gil Lee |
| 12:30 – 13:30 | Lunch Break (Doric Room) | |
| 13:30 – 14:00 | Lecture 4: Retrieval Augmented Generation (RAG) and Beyond: Boosting the Performance of LLMs | Prof Jae-Gil Lee |
| 14:00 – 15:00 | Lecture 5: Integrating Graphs and Large Language Models for Faithful Reasoning | Prof Shirui Pan |
| 15:00 – 15:30 | Tea Break (Doric Room) | |
| 15:30 – 17:00 | Panel: The impact of GAI on our research and education - What should we do now for a better future | Prof Huan Liu & Lecturers |
| 17:00 – 17.05 | Closing Remarks | Prof Longbing Cao |
| 18:00 – 20:00 | Reception (Marble Foyer, Level 1) | |
Prof. Huan Liu
Arizona State University, USA
Panel: The impact of GAI on our research and education - What should we do now for a better future
Bio
Dr. Huan Liu is a Regents professor of computer science and engineering with the School of Computing and Augmented Intelligence in the Ira A. Fulton Schools of Engineering. At Arizona State University, he was recognized for excellence in teaching and research in Computer Science and Engineering and received the 2014 President's Award for Innovation.
Prof. James Kwok
Hong Kong University of Science and Technology, China
Bio
James Kwok is a Professor in the Department of Computer Science and Engineering, Hong Kong University of Science and Technology. He is an IEEE Fellow.
Prof. Myra Spiliopoulou
Otto-von-Guericke University Magdeburg, Germany
Lecture: Temporal Learning in Healthcare - How to Care of the Missing Data
Abstract
The popularity of AI in healthcare is increasing, foremostly for tasks involving the analysis of images and signal, and for insight acquisition from Electronic Health Records (EHR). The temporal dimension of healthcare data is also often taken into account, whereby we must distinguish among the time scale of signal data (eg life signals in an Intensive Care Unit), the time scale of recordings during a patient's stay in the hospital and the time scale of patient monitoring over months or years. Missingness takes different forms among these scales, and we look at these forms in turn.
The solutions to temporal learning under missingness are conceptually simple to formulate: Ignore, Impute (without being able to verify), Actively ask an oracle (if we find one), Exploit (when the missingness is informative). We look at these options and discuss where they are applicable and what are their limitations.
Bio
MYRA SPILIOPOULOU is Professor of Business Information Systems at the Faculty of Computer Science, Otto-von-Guericke-University Magdeburg, Germany. Her main research is on mining dynamic complex data. Her publications are on mining complex streams, mining evolving objects, adapting models to drift and building models that capture drift.
Prof. Vincent Tseng
National Yang Ming Chiao Tung University, Taiwan
Bio
Vincent S. Tseng is currently a Chair Professor at Department of Computer Science in National Yang Ming Chiao Tung University (NYCU). He served as the founding director for Institute of Data Science and Engineering in NCTU during 2017-2020, chair for IEEE CIS Tainan Chapter during 2013-2015 and the president of Taiwanese Association for Artificial Intelligence during 2011-2012.
Prof. Jae-Gil Lee
Korea Advanced Institute of Science and Technology (KAIST), Korea
Lecture: Retrieval Augmented Generation (RAG) and Beyond: Boosting the Performance of LLMs
Abstract
Retrieval-Augmented Generation (RAG) has emerged as a powerful paradigm for overcoming key limitations of Large Language Models (LLMs). By integrating non-parametric retrieval modules with parametric generation, RAG enables models to dynamically ground their outputs in external knowledge bases, significantly reducing hallucination and improving performance on knowledge-intensive tasks. This lecture provides a deep dive into the RAG architecture, covering its foundational components. We will explore key advanced techniques such as query rewriting techniques and reranking strategies that refine both the retrieval and generation phases for improved end-to-end performance. Furthermore, we will examine cutting-edge developments including active retrieval and self-critique or self-knowledge. The practical relevance of RAG will be highlighted through its integration in production-level systems such as Perplexity. Last, we will also discuss the broader applicability of RAG across domains and its prospective evolution, including its role in agentic systems and real-time knowledge-grounded applications.
Bio
Professor Lee is a Professor at School of Computing, Korea Advanced Institute of Science and Technology (KAIST) and leads the Data Mining Lab. Before that, he was a Postdoctoral Researcher at IBM Almaden Research Center and a Postdoc Research Associate at Department of Computer Science, University of Illinois at Urbana-Champaign.
Prof. Shirui Pan
Griffith University, Australia
Lecture: Integrating Graphs and Large Language Models for Faithful Reasoning
Abstract
Large Language Models (LLMs) such as ChatGPT and Gemini have gained significant attention for their emergent abilities and generalizability. However, as black-box models, they face limitations in capturing and accessing factual knowledge. In contrast, graphs—both as data structures and mathematical modeling tools—offer rich factual information in a structured format, enhancing the inference capabilities and interpretability of LLMs. This lecture will introduce recent research on integrating graphs with LLMs for faithful reasoning. In particular, it will present a graph-enhanced LLM framework called Reasoning on Graphs (ROG), which leverages knowledge graphs to support accurate and interpretable reasoning. ROG follows a planning–retrieval–reasoning paradigm: first, it enables LLMs to generate a plan for retrieving relevant knowledge from a knowledge graph; then, based on the retrieved information, the LLM performs faithful reasoning.
Additionally, the lecture will highlight some recent advancements in graph foundation models, which facilitate zero-shot reasoning across new domains. The session will conclude with a brief discussion on future directions in this rapidly evolving field.
Bio
Shirui Pan is a Professor and an ARC Future Fellow with the School of Information and Communication Technology, Griffith University, Australia. He is a Co-Director of TrustAGI Lab. Before joining Griffith in 2022, he was Senior Lecturer (Associate Professor) with the Faculty of Information Technology, Monash University. He received his Ph.D degree in computer science from University of Technology Sydney (UTS), Australia.