Jia-Qi Lin
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Jia-Qi Lin has a total of 68 co-authors. Here are the top co-authors by number of shared publications:
- Chang-Dong Wang
Sun Yat-sen University
13 shared publications - Man-Sheng Chen
School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou, China
10 shared publications - Dong Huang
South China Agricultural University
4 shared publications - Haizhang Zhang
School of Mathematics (Zhuhai), Sun Yat-Sen University, Zhuhai 519082, China
3 shared publications - Li Li
Department of Emergency Medicine, Sun Yat-sen Memorial Hospital, Sun Yat-sen University, Guangzhou, China.
3 shared publications - Tao Yu
Department of Emergency Medicine, Sun Yat-sen Memorial Hospital, Sun Yat-sen University, Guangzhou, China.
3 shared publications
Latest Publications
- Open Access
- 28 May 2026
- Published by Springer Nature in Communications Medicine
- Vol. 6  (1)
Deep knowledge-driven multi-modal fusion for diagnosis and prognosis of SI-ARDS
Sepsis-Induced Acute Respiratory Distress Syndrome (SI-ARDS) presents significant diagnostic and prognostic challenges due to its complex clinical manifestations and high mortality rate. We developed a deep Knowledge-Driven multi-Modal Fusion (KDMF) framework for the accurate diagnosis and prognosis of SI-ARDS. The model leverages multi-modal data, including CT images, CT reports, and laboratory indicators, alongside a disease-specific knowledge graph. KDMF achieves superior performance in predicting SI-ARDS incidence (AUC 0.930) and time to 28-day mortality (AUC 0.843, C-index 0.833). Comprehensive error analysis and ablation studies demonstrate the critical contributions of each data modality and the integrated knowledge graph. The results highlight the potential of KDMF to enhance early intervention and treatment strategies, underscoring the robustness and interpretability of the framework in clinical applications.
- 28 May 2026
- Published by Springer Nature in Communications Medicine
- Vol. 6  (1)
Open AccessDeep knowledge-driven multi-modal fusion for diagnosis and prognosis of SI-ARDS
Sepsis-Induced Acute Respiratory Distress Syndrome (SI-ARDS) presents significant diagnostic and prognostic challenges due to its complex clinical manifestations and high mortality rate. We developed a deep Knowledge-Driven multi-Modal Fusion (KDMF) framework for the accurate diagnosis and prognosis of SI-ARDS. The model leverages multi-modal data, including CT images, CT reports, and laboratory indicators, alongside a disease-specific knowledge graph. KDMF achieves superior performance in predicting SI-ARDS incidence (AUC 0.930) and time to 28-day mortality (AUC 0.843, C-index 0.833). Comprehensive error analysis and ablation studies demonstrate the critical contributions of each data modality and the integrated knowledge graph. The results highlight the potential of KDMF to enhance early intervention and treatment strategies, underscoring the robustness and interpretability of the framework in clinical applications.
- Open Access
- 31 July 2025
- Published by Springer Nature in Communications Medicine
- Vol. 5  (1) , 1-15
A foundational triage system for improving accuracy in moderate acuity level emergency classifications
Background Triage is an essential part of Emergency Medicine, which may be assisted by AI models due to limited availability of medical staff. However, AI models for aiding triage have difficulty in identifying levels that are difficult or ambiguous for human clinicians to distinguish. This study aims to develop a more reliable triage model that improves the accuracy of classification, especially for cases with moderate acuity. Methods We developed a new triage model called KUTS, a foundational classification model for emergency triage, which leverages a knowledge prompt-tuning encoder and an uncertainty-based classifier. KUTS takes tabular data and chief complaints as input, and then assign different acuity levels to patients based on their condition acuity. We trained and tested the model on multiple real-world emergency department datasets. Results Here we show that on the most difficult level for human to distinguish (moderate acuity level), our KUTS substantially outperforms the previous shallow single-modal methods, deep single-modal methods and deep multi-modal methods on AUC score, by an average of 0.19, 0.35, and 0.13 (from 0.76, 0.60, 0.82 to 0.95), respectively. Besides, on all the triage levels, our KUTS also outperforms the previous shallow single-modal methods, deep single-modal methods and deep multi-modal methods on AUC score, by an average of 0.14, 0.20 and 0.06 (from 0.82, 0.76, 0.90 to 0.96), respectively. Conclusions KUTS provides a foundational framework and paradigm for the study of emergency triage, and facilitates the development of more efficient triage systems.
- 31 July 2025
- Published by Springer Nature in Communications Medicine
- Vol. 5  (1) , 1-15
Open AccessA foundational triage system for improving accuracy in moderate acuity level emergency classifications
Background Triage is an essential part of Emergency Medicine, which may be assisted by AI models due to limited availability of medical staff. However, AI models for aiding triage have difficulty in identifying levels that are difficult or ambiguous for human clinicians to distinguish. This study aims to develop a more reliable triage model that improves the accuracy of classification, especially for cases with moderate acuity. Methods We developed a new triage model called KUTS, a foundational classification model for emergency triage, which leverages a knowledge prompt-tuning encoder and an uncertainty-based classifier. KUTS takes tabular data and chief complaints as input, and then assign different acuity levels to patients based on their condition acuity. We trained and tested the model on multiple real-world emergency department datasets. Results Here we show that on the most difficult level for human to distinguish (moderate acuity level), our KUTS substantially outperforms the previous shallow single-modal methods, deep single-modal methods and deep multi-modal methods on AUC score, by an average of 0.19, 0.35, and 0.13 (from 0.76, 0.60, 0.82 to 0.95), respectively. Besides, on all the triage levels, our KUTS also outperforms the previous shallow single-modal methods, deep single-modal methods and deep multi-modal methods on AUC score, by an average of 0.14, 0.20 and 0.06 (from 0.82, 0.76, 0.90 to 0.96), respectively. Conclusions KUTS provides a foundational framework and paradigm for the study of emergency triage, and facilitates the development of more efficient triage systems.
- 31 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks and Learning Systems
- Vol. 36  (8) , 14678-14690
Contrastive Ensemble Clustering
Ensemble clustering aims to combine different base clusterings into a better clustering than that of the individual one. In general, a co-association matrix depicting the pairwise affinity between different data samples is constructed by average fusion or weighted fusion of the connective matrices from multiple base clusterings. Despite the significant success, the existing works fail to capture the global structure information from multiple noisy connective matrices. Meanwhile, the locality property of the resulting representation matrix could not be explicitly preserved. In this article, we propose a novel contrastive ensemble clustering (CEC) method. Specifically, a consensus mapping model is designed for the discovery of the latent representation from the noisy observations with distinct confidences. Furthermore, a contrastive regularizer is dexterously formulated to refine the latent representation while preserving its locality property. Extensive experiments conducted on several benchmark datasets demonstrate the superiority of the proposed CEC method. To the best of our knowledge, it is the first time to explore the potential of latent representation learning and contrastive components for the ensemble clustering task.
- 31 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks and Learning Systems
- Vol. 36  (8) , 14678-14690
Contrastive Ensemble Clustering
Ensemble clustering aims to combine different base clusterings into a better clustering than that of the individual one. In general, a co-association matrix depicting the pairwise affinity between different data samples is constructed by average fusion or weighted fusion of the connective matrices from multiple base clusterings. Despite the significant success, the existing works fail to capture the global structure information from multiple noisy connective matrices. Meanwhile, the locality property of the resulting representation matrix could not be explicitly preserved. In this article, we propose a novel contrastive ensemble clustering (CEC) method. Specifically, a consensus mapping model is designed for the discovery of the latent representation from the noisy observations with distinct confidences. Furthermore, a contrastive regularizer is dexterously formulated to refine the latent representation while preserving its locality property. Extensive experiments conducted on several benchmark datasets demonstrate the superiority of the proposed CEC method. To the best of our knowledge, it is the first time to explore the potential of latent representation learning and contrastive components for the ensemble clustering task.
- 26 September 2024
- Published by Elsevier in Neural Networks
- Vol. 181, 106755
Interaction-knowledge semantic alignment for recommendation
- 26 September 2024
- Published by Elsevier in Neural Networks
- Vol. 181, 106755
Interaction-knowledge semantic alignment for recommendation
- Open Access
- 25 July 2024
- Published by Springer Nature in Research Square
KUTS: A Foundational Triage System for Improving Accuracy in Mid-Level Emergency Classifications
Triage is of great importance in the procedure of the emergency diagnosis, which needs artificial intelligence to assist due to the scarcity of medical resources. However, AI models for aiding triage have difficulty in identifying levels that are difficult or ambiguous for human clinicians to distinguish. Here, to address this challenge and improve the performance of triage models, we propose KUTS, a foundational classification model for emergency triage, which leverages a knowledge prompt-tuning encoder and an uncertainty-based classifier. KUTS takes tabular data and chief complaints as input, and then generates the recommended triage level for human clinicians. On the most difficult level for human to distinguish (Level 3), our KUTS significantly outperformed the previous shallow single-modal methods, deep single-modal methods and deep multi-modal methods by an average of 14.85%, 28.38% and 11.14%, respectively. Besides, on all the triage levels, our KUTS also outperformed the previous shallow single-modal methods, deep single-modal methods and deep multi-modal methods by an average of 9.02%, 9.45% and 4.18%, respectively. KUTS provides a foundational framework and paradigm for the study of emergency triage, and can be extended to scenarios that require pre-hospital emergency treatment such as disasters and accidents.
- 25 July 2024
- Published by Springer Nature in Research Square
Open AccessKUTS: A Foundational Triage System for Improving Accuracy in Mid-Level Emergency Classifications
Triage is of great importance in the procedure of the emergency diagnosis, which needs artificial intelligence to assist due to the scarcity of medical resources. However, AI models for aiding triage have difficulty in identifying levels that are difficult or ambiguous for human clinicians to distinguish. Here, to address this challenge and improve the performance of triage models, we propose KUTS, a foundational classification model for emergency triage, which leverages a knowledge prompt-tuning encoder and an uncertainty-based classifier. KUTS takes tabular data and chief complaints as input, and then generates the recommended triage level for human clinicians. On the most difficult level for human to distinguish (Level 3), our KUTS significantly outperformed the previous shallow single-modal methods, deep single-modal methods and deep multi-modal methods by an average of 14.85%, 28.38% and 11.14%, respectively. Besides, on all the triage levels, our KUTS also outperformed the previous shallow single-modal methods, deep single-modal methods and deep multi-modal methods by an average of 9.02%, 9.45% and 4.18%, respectively. KUTS provides a foundational framework and paradigm for the study of emergency triage, and can be extended to scenarios that require pre-hospital emergency treatment such as disasters and accidents.
- 12 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Emerging Topics in Computational Intelligence
- Vol. 8  (5) , 3534-3543
Sparse Graph Tensor Learning for Multi-View Spectral Clustering
Multi-view spectral clustering has achieved impressive performance by learning multiple robust and meaningful similarity graphs for clustering. Generally, the existing literatures often construct multiple similarity graphs by certain similarity measure (e.g. the Euclidean distance), which lack the desired ability to learn sparse and reliable connections that carry critical information in graph learning while preserving the low-rank structure. Regarding the challenges, a novel Sparse Graph Tensor Learning for Multi-view Spectral Clustering (SGTL) method is designed in this paper, where multiple similarity graphs are seamlessly coupled with the cluster indicators and constrained with a low-rank graph tensor. Specifically, a novel graph learning paradigm is designed by establishing an explicit theoretical connection between the similarity matrices and the cluster indicator matrices, in order that the constructed similarity graphs enjoy the desired block diagonal and sparse property for learning a small portion of reliable links. Then, we stack multiple similarity matrices into a low-rank graph tensor to better preserve the low-rank structure of the reliable links in graph learning, where the key knowledge conveyed by singular values from different views is explicitly considered. Extensive experiments on several benchmark datasets demonstrate the superiority of SGTL.
- 12 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Emerging Topics in Computational Intelligence
- Vol. 8  (5) , 3534-3543
Sparse Graph Tensor Learning for Multi-View Spectral Clustering
Multi-view spectral clustering has achieved impressive performance by learning multiple robust and meaningful similarity graphs for clustering. Generally, the existing literatures often construct multiple similarity graphs by certain similarity measure (e.g. the Euclidean distance), which lack the desired ability to learn sparse and reliable connections that carry critical information in graph learning while preserving the low-rank structure. Regarding the challenges, a novel Sparse Graph Tensor Learning for Multi-view Spectral Clustering (SGTL) method is designed in this paper, where multiple similarity graphs are seamlessly coupled with the cluster indicators and constrained with a low-rank graph tensor. Specifically, a novel graph learning paradigm is designed by establishing an explicit theoretical connection between the similarity matrices and the cluster indicator matrices, in order that the constructed similarity graphs enjoy the desired block diagonal and sparse property for learning a small portion of reliable links. Then, we stack multiple similarity matrices into a low-rank graph tensor to better preserve the low-rank structure of the reliable links in graph learning, where the key knowledge conveyed by singular values from different views is explicitly considered. Extensive experiments on several benchmark datasets demonstrate the superiority of SGTL.
- 30 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks and Learning Systems
- Vol. 36  (4) , 6466-6477
Dual Information Enhanced Multiview Attributed Graph Clustering
Multiview attributed graph clustering is an important approach to partition multiview data based on the attribute characteristics and adjacent matrices from different views. Some attempts have been made in using graph neural network (GNN), which have achieved promising clustering performance. Despite this, few of them pay attention to the inherent specific information embedded in multiple views. Meanwhile, they are incapable of recovering the latent high-level representation from the low-level ones, greatly limiting the downstream clustering performance. To fill these gaps, a novel dual information enhanced multiview attributed graph clustering (DIAGC) method is proposed in this article. Specifically, the proposed method introduces the specific information reconstruction (SIR) module to disentangle the explorations of the consensus and specific information from multiple views, which enables graph convolutional network (GCN) to capture the more essential low-level representations. Besides, the contrastive learning (CL) module maximizes the agreement between the latent high-level representation and low-level ones and enables the high-level representation to satisfy the desired clustering structure with the help of the self-supervised clustering (SC) module. Extensive experiments on several real-world benchmarks demonstrate the effectiveness of the proposed DIAGC method compared with the state-of-the-art baselines.
- 30 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks and Learning Systems
- Vol. 36  (4) , 6466-6477
Dual Information Enhanced Multiview Attributed Graph Clustering
Multiview attributed graph clustering is an important approach to partition multiview data based on the attribute characteristics and adjacent matrices from different views. Some attempts have been made in using graph neural network (GNN), which have achieved promising clustering performance. Despite this, few of them pay attention to the inherent specific information embedded in multiple views. Meanwhile, they are incapable of recovering the latent high-level representation from the low-level ones, greatly limiting the downstream clustering performance. To fill these gaps, a novel dual information enhanced multiview attributed graph clustering (DIAGC) method is proposed in this article. Specifically, the proposed method introduces the specific information reconstruction (SIR) module to disentangle the explorations of the consensus and specific information from multiple views, which enables graph convolutional network (GCN) to capture the more essential low-level representations. Besides, the contrastive learning (CL) module maximizes the agreement between the latent high-level representation and low-level ones and enables the high-level representation to satisfy the desired clustering structure with the help of the self-supervised clustering (SC) module. Extensive experiments on several real-world benchmarks demonstrate the effectiveness of the proposed DIAGC method compared with the state-of-the-art baselines.
- 29 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks and Learning Systems
- Vol. 36  (4) , 7184-7195
Contrastive Multiview Attribute Graph Clustering With Adaptive Encoders
Multiview attribute graph clustering aims to cluster nodes into disjoint categories by taking advantage of the multiview topological structures and the node attribute values. However, the existing works fail to explicitly discover the inherent relationships in multiview topological graph matrices while considering different properties between the graphs. Besides, they cannot well handle the sparse structure of some graphs in the learning procedure of graph embeddings. Therefore, in this article, we propose a novel contrastive multiview attribute graph clustering (CMAGC) with adaptive encoders method. Within this framework, the adaptive encoders concerning different properties of distinct topological graphs are chosen to integrate multiview attribute graph information by checking whether there exists high-order neighbor information or not. Meanwhile, the number of layers of the GCN encoders is selected according to the prior knowledge related to the characteristics of different topological graphs. In particular, the feature-level and cluster-level contrastive learning are conducted on the multiview soft assignment representations, where the union of the first-order neighbors from the corresponding graph pairs is regarded as the positive pairs for data augmentation and the sparse neighbor information problem in some graphs can be well dealt with. To the best of our knowledge, it is the first time to explicitly deal with the inherent relationships from the interview and intraview perspectives. Extensive experiments are conducted on several datasets to verify the superiority of the proposed CMAGC method compared with the state-of-the-art methods.
- 29 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks and Learning Systems
- Vol. 36  (4) , 7184-7195
Contrastive Multiview Attribute Graph Clustering With Adaptive Encoders
Multiview attribute graph clustering aims to cluster nodes into disjoint categories by taking advantage of the multiview topological structures and the node attribute values. However, the existing works fail to explicitly discover the inherent relationships in multiview topological graph matrices while considering different properties between the graphs. Besides, they cannot well handle the sparse structure of some graphs in the learning procedure of graph embeddings. Therefore, in this article, we propose a novel contrastive multiview attribute graph clustering (CMAGC) with adaptive encoders method. Within this framework, the adaptive encoders concerning different properties of distinct topological graphs are chosen to integrate multiview attribute graph information by checking whether there exists high-order neighbor information or not. Meanwhile, the number of layers of the GCN encoders is selected according to the prior knowledge related to the characteristics of different topological graphs. In particular, the feature-level and cluster-level contrastive learning are conducted on the multiview soft assignment representations, where the union of the first-order neighbors from the corresponding graph pairs is regarded as the positive pairs for data augmentation and the sparse neighbor information problem in some graphs can be well dealt with. To the best of our knowledge, it is the first time to explicitly deal with the inherent relationships from the interview and intraview perspectives. Extensive experiments are conducted on several datasets to verify the superiority of the proposed CMAGC method compared with the state-of-the-art methods.
- 26 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks and Learning Systems
- Vol. 35  (9) , 11706-11719
Motif-Based Contrastive Learning for Community Detection
Community detection has become a prominent task in complex network analysis. However, most of the existing methods for community detection only focus on the lower order structure at the level of individual nodes and edges and ignore the higher order connectivity patterns that characterize the fundamental building blocks within the network. In recent years, researchers have shown interest in motifs and their role in network analysis. However, most of the existing higher order approaches are based on shallow methods, failing to capture the intricate nonlinear relationships between nodes. In order to better fuse higher order and lower order structural information, a novel deep learning framework called motif-based contrastive learning for community detection (MotifCC) is proposed. First, a higher order network is constructed based on motifs. Subnetworks are then obtained by removing isolated nodes, addressing the fragmentation issue in the higher order network. Next, the concept of contrastive learning is applied to effectively fuse various kinds of information from nodes, edges, and higher order and lower order structures. This aims to maximize the similarity of corresponding node information, while distinguishing different nodes and different communities. Finally, based on the community structure of subnetworks, the community labels of all nodes are obtained by using the idea of label propagation. Extensive experiments on real-world datasets validate the effectiveness of MotifCC.
- 26 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks and Learning Systems
- Vol. 35  (9) , 11706-11719
Motif-Based Contrastive Learning for Community Detection
Community detection has become a prominent task in complex network analysis. However, most of the existing methods for community detection only focus on the lower order structure at the level of individual nodes and edges and ignore the higher order connectivity patterns that characterize the fundamental building blocks within the network. In recent years, researchers have shown interest in motifs and their role in network analysis. However, most of the existing higher order approaches are based on shallow methods, failing to capture the intricate nonlinear relationships between nodes. In order to better fuse higher order and lower order structural information, a novel deep learning framework called motif-based contrastive learning for community detection (MotifCC) is proposed. First, a higher order network is constructed based on motifs. Subnetworks are then obtained by removing isolated nodes, addressing the fragmentation issue in the higher order network. Next, the concept of contrastive learning is applied to effectively fuse various kinds of information from nodes, edges, and higher order and lower order structures. This aims to maximize the similarity of corresponding node information, while distinguishing different nodes and different communities. Finally, based on the community structure of subnetworks, the community labels of all nodes are obtained by using the idea of label propagation. Extensive experiments on real-world datasets validate the effectiveness of MotifCC.
- Open Access
- 26 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 3069-3077
On Regularizing Multiple Clusterings for Ensemble Clustering by Graph Tensor Learning
Ensemble clustering has shown its promising ability in fusing multiple base clusterings into a probably better and more robust clustering result. Typically, the co-association matrix based ensemble clustering methods attempt to integrate multiple connective matrices from base clusterings by weighted fusion to acquire a common graph representation. However, few of them are aware of the potential noise or corruption from the common representation by direct integration of different connective matrices with distinct cluster structures, and further consider the mutual information propagation between the input observations. In this paper, we propose a Graph Tensor Learning based Ensemble Clustering (GTLEC) method to refine multiple connective matrices by the substantial rank recovery and graph tensor learning. Within this framework, each input connective matrix is dexterously refined to approximate a graph structure by obeying the theoretical rank constraint with an adaptive weight coefficient. Further, we stack multiple refined connective matrices into a three-order tensor to extract their higher-order similarities via graph tensor learning, where the mutual information propagation across different graph matrices will also be promoted. Extensive experiments on several challenging datasets have confirmed the superiority of GTLEC compared with the state-of-the-art.
- 26 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 3069-3077
Open AccessOn Regularizing Multiple Clusterings for Ensemble Clustering by Graph Tensor Learning
Ensemble clustering has shown its promising ability in fusing multiple base clusterings into a probably better and more robust clustering result. Typically, the co-association matrix based ensemble clustering methods attempt to integrate multiple connective matrices from base clusterings by weighted fusion to acquire a common graph representation. However, few of them are aware of the potential noise or corruption from the common representation by direct integration of different connective matrices with distinct cluster structures, and further consider the mutual information propagation between the input observations. In this paper, we propose a Graph Tensor Learning based Ensemble Clustering (GTLEC) method to refine multiple connective matrices by the substantial rank recovery and graph tensor learning. Within this framework, each input connective matrix is dexterously refined to approximate a graph structure by obeying the theoretical rank constraint with an adaptive weight coefficient. Further, we stack multiple refined connective matrices into a three-order tensor to extract their higher-order similarities via graph tensor learning, where the mutual information propagation across different graph matrices will also be promoted. Extensive experiments on several challenging datasets have confirmed the superiority of GTLEC compared with the state-of-the-art.