Generalizable and Efficient Transfer Learning for Graph and Language Models

The Hong Kong University of Science and Technology (Guangzhou)

Data Science and Analytics Thrust

PhD Thesis Examination

By Mr. Jiashun CHENG

ABSTRACT

Transfer learning, as a key paradigm in modern machine learning, has rapidly advanced the scalability and effectiveness of model deployment by enabling knowledge reuse across tasks. This thesis identifies three principal challenges that arise at different stages of its evolving landscape: (1) generalizable graph pre-training, (2) effective adaptation for graphs, and (3) efficient adaptation for large language models (LLMs), and presents three innovative approaches to tackle these challenges.

Our first framework, WGDN, revisits the long-overlooked generative graph pre-training by proposing the augmentation-adaptive graph Wiener filter that better captures the topological nature of graphs within the reconstruction paradigm, thereby improving the generalizability of graph representation learning. Subsequently, PAF addresses the challenges of limited supervision and structural heterophily in graph anomaly detection by introducing a unified framework that combines multi-filter spectral pre-training with adaptive fine-tuning, enabling task-aware and effective adaptation. Furthermore, SeLoRA targets the parameter redundancy in Low-Rank Adaptation (LoRA) fine-tuning for LLMs, and introduces a sparse spectral re-parameterization module that achieves highly efficient adaptation while preserving expressive capacity.

Extensive experiments across diverse benchmarks and downstream tasks in both graph and language domains validate the effectiveness and efficiency of these approaches. Collectively, they aim to advance the frontier of transfer learning toward greater generalizability, adaptability, and resource efficiency.

TEC

Chairperson: Prof Sihong XIE
Prime Supervisor: Prof Fugee TSUNG
Co-Supervisor: Prof Jia LI
Examiners:
Prof Qiong LUO
Prof Lei ZHU
Prof Yingcong CHEN
Prof Yang YANG