Export
Jia-Qi Lin
· 4 previous affiliations
- Researcher · Since 2025
- Sun Yat-sen University2021–2025
- PhD student · 2021–2025
- Fuzhou University2018–2021
- MS student · 2018–2021
- Taiyuan University of Technology2014–2018
- Undergrad student · 2014–2018
Research Overview
Jia-Qi Lin focuses on multi-view clustering, graph tensor learning, and data fusion in unsupervised machine learning. Early work addressed the integration of multiple base clusterings and the challenges of uncoupled or incomplete multi-view data. More recently, this trajectory has expanded to include semantic alignment for knowledge-graph-enhanced recommender systems and higher-order community detection.
Research Trajectory
Jia-Qi Lin’s research initially centered on improving the robustness and scalability of clustering algorithms, particularly when dealing with complex, multi-source data. A recurring challenge in this domain is how to effectively fuse multiple representations without amplifying noise. In 2023, Lin was the second author on On Regularizing Multiple Clusterings for Ensemble Clustering by Graph Tensor Learning. This work addressed the limitations of traditional co-association matrix methods in ensemble clustering, proposing a graph tensor learning framework to mitigate noise and better capture mutual information between input observations. This research aligns with the broader area of Consensus Graph Learning for Multi-View Data Fusion, where the goal is to extract a reliable unified representation from disparate data sources.
In 2024, Lin led significant work addressing the practical limitations of multi-view learning in real-world scenarios, specifically focusing on data that is both incomplete and uncoupled. Traditional multi-view algorithms often assume perfect synchronization between views, a constraint that frequently fails in practice. As first author on Incomplete Data Meets Uncoupled Case: A Challenging Task of Multiview Clustering, Lin tackled the dual problem of missing data and unclear cross-view correlations. This was closely followed by another first-author contribution, A Tensor Approach for Uncoupled Multiview Clustering, which utilized tensor methods to handle features that are not strictly aligned across different views. These papers represent a core contribution to the research direction of Multi-View Learning without Instance Correspondence and advance themes such as Tensorized Multi-view Representation Learning and Disentangling Shared and View Specific Features in Multi View Learning.
Alongside this primary focus on uncoupled multi-view clustering, Lin has participated as a co-author in related graph learning and network analysis projects. As a middle author, Lin contributed to Sparse Graph Tensor Learning for Multi-View Spectral Clustering (2024), which explored learning sparse, reliable connections for spectral clustering while preserving low-rank structures. In the same year, Lin also collaborated on Motif-Based Contrastive Learning for Community Detection, a project that moved beyond node-level structures to leverage higher-order connectivity patterns (motifs) for analyzing complex networks.
By 2025, Lin’s focus on graph-based data fusion extended into the domain of recommender systems. Serving as second author on Interaction-knowledge semantic alignment for recommendation, Lin helped develop methods to integrate knowledge graphs with user-item interaction data. This work specifically addressed the semantic gap between structural knowledge and behavioral data, ensuring that fused representations account for personalized user preferences rather than just item characteristics.
Current Directions
The most recent papers in the evidence indicate a shift from purely unsupervised multi-view clustering toward applied graph representation learning, particularly in recommendation systems and complex network analysis. Current work investigates how to bridge the semantic differences between distinct types of graph structures—such as behavioral interaction graphs and factual knowledge graphs. By focusing on semantic alignment and higher-order structural motifs, Lin's recent research seeks to build more robust, context-aware representations that can alleviate data sparsity and improve personalization in predictive models.
Profile updated
| Title | Venue | Year |
|---|---|---|
| Empowering Test-Time Adaptation with Complementary Vision-Language Knowledge in Open-World Scenarios | Openreview | 2026 |
| Interaction-knowledge semantic alignment for recommendation | Neural Networks | 2025 |
| Incomplete Data Meets Uncoupled Case: A Challenging Task of Multiview Clustering | IEEE TNNLS | 2024 |
| A Tensor Approach for Uncoupled Multiview Clustering | IEEE Trans. Cybernetics | 2024 |
| Motif-Based Contrastive Learning for Community Detection | IEEE TNNLS | 2024 |
| Sparse Graph Tensor Learning for Multi-View Spectral Clustering | IEEE Trans. Emerg. Top. Comput. Intell | 2024 |
| On Regularizing Multiple Clusterings for Ensemble Clustering by Graph Tensor Learning | ACM Multimedia | 2023 |
Loading new papers…
Loading research neighbors…
See something wrong? Report a mistake