About
Hi Kuen Yu is a PhD Student in the ASI Lab. of Department of Computer Science, City University of Hong Kong.
Hi Kuen Yu works on software log analysis and anomaly detection, focusing on making deep models robust to heterogeneous log formats and labeled-data scarcity. In LogMeta, a semi-supervised MAML-based framework with a hybrid RoBERTa plus Bi-LSTM/attention model achieves up to a 28.3% F1 improvement over prior methods in low-resource scenarios. Through a task-oriented evaluation of parsing and detection across industrial datasets, he shows that models strong on benchmarks degrade substantially on real industrial logs, and that improved parsing only benefits anomaly detection when corrections preserve anomaly-critical semantic cues. In LogRoBERTa, he demonstrates a parser-free hybrid model using DPP-based selection that outperforms state-of-the-art including fully supervised baselines on four datasets while staying robust on low-resource data.
Research Interests
Publications (15)
-
COMPSAC2024