Liming Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Liming Wang has a total of 6 co-authors. Here are the top co-authors by number of shared publications:
- Hong Xie
School of Computer Science and Technology, University of Science and Technology of China
1 shared publication - Jian Tan
Alibaba Cloud Computing, Sunnyvale, CA, USA
1 shared publication - John Chi-Shing Lui
The Chinese University of Hong Kong (CUHK)
1 shared publication - Meihui Zhang
Beijing Institute of Technology
1 shared publication - Ye Li
Alibaba Cloud Computing, Hangzhou, China
1 shared publication - Zhaojing Luo
Beijing Institute of Technology, Beijing, China
1 shared publication
Latest Publications
- Open Access
- 4 December 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Management of Data
- Vol. 3 (6) , 1-26
AixelNet: A Pre-trained Model with Table-aware Adaptation for Structured Data Prediction
Structured data prediction supports key applications in healthcare, finance, e-commerce, among others. As structured data becomes increasingly heterogeneous and complex, there is a growing need for scalable models that can generalize across diverse tables. Traditional machine learning and deep learning models for structured data often have task-specific architectures requiring computationally retraining for a new prediction task. Consequently, recent studies explore pre-training for structured data, which requires no additional training for a new task. While pre-training has achieved great success in NLP and CV, the research of pre-training for structured data remains preliminary because structured data has its unique characteristics, e.g., complicated correlations and dependencies between features. Existing studies adopt a fixed pre-trained model architecture and less attention has been paid to table-specific characteristics. This limits their ability to adapt to new prediction tasks. To address this challenge, we propose AixelNet, a pre-trained model for structured data prediction that supports table-aware adaptation. We first design a meta feature extraction module that summarizes table-level characteristics enabling model customization. Instead of relying on a single model, AixelNet adopts a multi-model framework with multiple base predictors to capture diverse feature interaction patterns across tables. Finally, these base predictor models are dynamically assigned to different tables for prediction based on the extracted table meta features via a designed hypernetwork, enabling flexible and table-aware model composition for different tables. To further improve generalization and efficiency, we design regularization methods to encourage balanced predictor model usage and diversified learned representations, as well as design a sparse update strategy for sparsely updating relevant predictor models during pre-training. Extensive experiments on 20 classification and 20 regression tasks on tables confirm AixelNet's effectiveness in improving predictive performance compared to eight state-of-the-art baselines and demonstrate its efficiency.
- 4 December 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Management of Data
- Vol. 3 (6) , 1-26
Open AccessAixelNet: A Pre-trained Model with Table-aware Adaptation for Structured Data Prediction
Structured data prediction supports key applications in healthcare, finance, e-commerce, among others. As structured data becomes increasingly heterogeneous and complex, there is a growing need for scalable models that can generalize across diverse tables. Traditional machine learning and deep learning models for structured data often have task-specific architectures requiring computationally retraining for a new prediction task. Consequently, recent studies explore pre-training for structured data, which requires no additional training for a new task. While pre-training has achieved great success in NLP and CV, the research of pre-training for structured data remains preliminary because structured data has its unique characteristics, e.g., complicated correlations and dependencies between features. Existing studies adopt a fixed pre-trained model architecture and less attention has been paid to table-specific characteristics. This limits their ability to adapt to new prediction tasks. To address this challenge, we propose AixelNet, a pre-trained model for structured data prediction that supports table-aware adaptation. We first design a meta feature extraction module that summarizes table-level characteristics enabling model customization. Instead of relying on a single model, AixelNet adopts a multi-model framework with multiple base predictors to capture diverse feature interaction patterns across tables. Finally, these base predictor models are dynamically assigned to different tables for prediction based on the extracted table meta features via a designed hypernetwork, enabling flexible and table-aware model composition for different tables. To further improve generalization and efficiency, we design regularization methods to encourage balanced predictor model usage and diversified learned representations, as well as design a sparse update strategy for sparsely updating relevant predictor models during pre-training. Extensive experiments on 20 classification and 20 regression tasks on tables confirm AixelNet's effectiveness in improving predictive performance compared to eight state-of-the-art baselines and demonstrate its efficiency.
- Open Access
- 30 April 2023
- Published by Association for Computing Machinery (ACM)
- p. 1960-1968
Interactive Log Parsing via Light-weight User Feedback
Template mining is one of the foundational tasks to support log analysis, which supports the diagnosis and troubleshooting of large scale Web applications. This paper develops a human-in-the-loop template mining framework to support interactive log analysis, which is highly desirable in real-world diagnosis or troubleshooting of Web applications but yet previous template mining algorithms fail to support it. We formulate three types of light-weight user feedback and based on them we design three atomic human-in-the-loop template mining algorithms. We derive mild conditions under which the outputs of our proposed algorithms are provably correct. We also derive upper bounds on the computational complexity and query complexity of each algorithm. We demonstrate the versatility of our proposed algorithms by combining them to improve the template mining accuracy of five representative algorithms over sixteen widely used benchmark datasets.
- 30 April 2023
- Published by Association for Computing Machinery (ACM)
- p. 1960-1968
Open AccessInteractive Log Parsing via Light-weight User Feedback
Template mining is one of the foundational tasks to support log analysis, which supports the diagnosis and troubleshooting of large scale Web applications. This paper develops a human-in-the-loop template mining framework to support interactive log analysis, which is highly desirable in real-world diagnosis or troubleshooting of Web applications but yet previous template mining algorithms fail to support it. We formulate three types of light-weight user feedback and based on them we design three atomic human-in-the-loop template mining algorithms. We derive mild conditions under which the outputs of our proposed algorithms are provably correct. We also derive upper bounds on the computational complexity and query complexity of each algorithm. We demonstrate the versatility of our proposed algorithms by combining them to improve the template mining accuracy of five representative algorithms over sixteen widely used benchmark datasets.