Huayu Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Huayu Li has a total of 38 co-authors. Here are the top co-authors by number of shared publications:
- Peiying Zhang
China University of Petroleum (East China)
3 shared publications - Xiangling Meng
School of Computer and Communication, University of Science and Technology Beijing, Beijing 100083, China
3 shared publications - Cuicui Wang
Qingdao Institute of Software, College of Computer Science and Technology, China University of Petroleum (East China), Qingdao, China
2 shared publications - Jing Liu
School of Cultural Heritage and Information Management, Shanghai University, Shanghai 200444, China
2 shared publications - Kai Liu
Tsinghua University
2 shared publications - Xiao He
School of Computer and Communication Engineering, University of Science and Technology Beijing, Beijing, China
2 shared publications
Latest Publications
- Open Access
- 1 September 2026
- Published by MDPI AG in Computation
- Vol. 14 (9) , 201
A Schur-Consistent GPU CPRW–AMG Preconditioner for Well-Coupled Fully Implicit Multiphase Flow Simulation
Fully implicit simulation of multiphase flow in porous media requires repeated solution of large, sparse, nonsymmetric Jacobian systems, where linear solvers and preconditioners often dominate the computational cost. In field-scale models with strong well controls and well–reservoir coupling, bottom-hole pressure (BHP) constraints may introduce slowly converging pressure-related error modes that are not fully represented by pressure-only constrained pressure residual (CPR) preconditioning. This paper develops a Schur-consistent graphics processing unit (GPU) implementation of constrained pressure residual with wells and algebraic multigrid (CPRW-AMG) for the Open Porous Media (OPM) Flow simulator. The method augments the pressure coarse space with one auxiliary BHP unknown per active well while preserving a reservoir-only fine-level Krylov system. The pressure–BHP coarse operator is constructed from the same assembled effective reservoir Jacobian used by the Krylov matrix–vector product and the GPU diagonal incomplete lower–upper (DILU) smoother, avoiding inconsistent treatment of well contributions between fine and coarse levels. A GPU-resident sparse update strategy refreshes well-related coarse entries, and the NVIDIA AMGX library is used for the scalar coarse solve. Benchmarks on SPE9, SPE10, Sleipner, and Norne cover small well-coupled, highly heterogeneous carbon dioxide storage and field-realistic reservoir models. Compared with CPU CPRW-AMG, the GPU implementation reduces total runtime by factors of 1.69–8.78 and achieves linear-solve speedups of up to 17.96×. Compared with GPU CPR-AMG, the performance benefit is case-dependent, indicating that CPRW-AMG is most effective when the reduction in well-induced slow error modes compensates for the additional coarse-level cost.
- 1 September 2026
- Published by MDPI AG in Computation
- Vol. 14 (9) , 201
Open AccessA Schur-Consistent GPU CPRW–AMG Preconditioner for Well-Coupled Fully Implicit Multiphase Flow Simulation
Fully implicit simulation of multiphase flow in porous media requires repeated solution of large, sparse, nonsymmetric Jacobian systems, where linear solvers and preconditioners often dominate the computational cost. In field-scale models with strong well controls and well–reservoir coupling, bottom-hole pressure (BHP) constraints may introduce slowly converging pressure-related error modes that are not fully represented by pressure-only constrained pressure residual (CPR) preconditioning. This paper develops a Schur-consistent graphics processing unit (GPU) implementation of constrained pressure residual with wells and algebraic multigrid (CPRW-AMG) for the Open Porous Media (OPM) Flow simulator. The method augments the pressure coarse space with one auxiliary BHP unknown per active well while preserving a reservoir-only fine-level Krylov system. The pressure–BHP coarse operator is constructed from the same assembled effective reservoir Jacobian used by the Krylov matrix–vector product and the GPU diagonal incomplete lower–upper (DILU) smoother, avoiding inconsistent treatment of well contributions between fine and coarse levels. A GPU-resident sparse update strategy refreshes well-related coarse entries, and the NVIDIA AMGX library is used for the scalar coarse solve. Benchmarks on SPE9, SPE10, Sleipner, and Norne cover small well-coupled, highly heterogeneous carbon dioxide storage and field-realistic reservoir models. Compared with CPU CPRW-AMG, the GPU implementation reduces total runtime by factors of 1.69–8.78 and achieves linear-solve speedups of up to 17.96×. Compared with GPU CPR-AMG, the performance benefit is case-dependent, indicating that CPRW-AMG is most effective when the reduction in well-induced slow error modes compensates for the additional coarse-level cost.
- Open Access
- 14 August 2026
- Published by MDPI AG in Electronics
- Vol. 15 (16) , 3633
HGCRec: A Heterogeneous Graph Contrastive Learning Framework for Cold-Start Multi-Modal Recommendation in Intelligent Information Systems
Personalized recommendation has become an essential component of intelligent information systems and electronic multimedia platforms. However, cold-start items with limited user–item interactions remain difficult to model, especially when collaborative signals are sparse and heterogeneous side information is underutilized. To address this problem, this paper proposes Heterogeneous Graph Contrastive Recommendation (HGCRec), a heterogeneous graph contrastive learning framework for cold-start multi-modal recommendation. HGCRec constructs a heterogeneous structural graph containing users, items, categories, brands, and semantic entities, while visual and textual information is incorporated as item-side feature views. A relation-aware heterogeneous graph encoder captures relation-specific structural semantics, and a cross-modal graph contrastive learning objective coordinates structural, visual, and textual item representations. An interaction-sparsity-aware weighting strategy allocates stronger contrastive supervision to items with fewer training interactions. Furthermore, an item-specific adaptive fusion module integrates collaborative, structural, visual, and textual representations according to interaction sparsity and learned multi-source representation states. Experiments are conducted on three Amazon multi-modal recommendation datasets, including Baby, Sports, and Clothing. The results show that HGCRec consistently outperforms representative graph-based and multi-modal recommendation baselines. Compared with the strongest baseline, Freezing and Denoising Graph Structures for Multimodal Recommendation (FREEDOM), HGCRec improves Recall@20 by 9.65%, 10.44%, and 12.00% on Baby, Sports, and Clothing, respectively, and improves NDCG@20 by 11.14%, 11.35%, and 13.10%. Sparsity-aware analysis further shows larger relative improvements for items with extremely limited training interactions, demonstrating the effectiveness of HGCRec under the evaluated interaction-sparse and few-shot settings.
- 14 August 2026
- Published by MDPI AG in Electronics
- Vol. 15 (16) , 3633
Open AccessHGCRec: A Heterogeneous Graph Contrastive Learning Framework for Cold-Start Multi-Modal Recommendation in Intelligent Information Systems
Personalized recommendation has become an essential component of intelligent information systems and electronic multimedia platforms. However, cold-start items with limited user–item interactions remain difficult to model, especially when collaborative signals are sparse and heterogeneous side information is underutilized. To address this problem, this paper proposes Heterogeneous Graph Contrastive Recommendation (HGCRec), a heterogeneous graph contrastive learning framework for cold-start multi-modal recommendation. HGCRec constructs a heterogeneous structural graph containing users, items, categories, brands, and semantic entities, while visual and textual information is incorporated as item-side feature views. A relation-aware heterogeneous graph encoder captures relation-specific structural semantics, and a cross-modal graph contrastive learning objective coordinates structural, visual, and textual item representations. An interaction-sparsity-aware weighting strategy allocates stronger contrastive supervision to items with fewer training interactions. Furthermore, an item-specific adaptive fusion module integrates collaborative, structural, visual, and textual representations according to interaction sparsity and learned multi-source representation states. Experiments are conducted on three Amazon multi-modal recommendation datasets, including Baby, Sports, and Clothing. The results show that HGCRec consistently outperforms representative graph-based and multi-modal recommendation baselines. Compared with the strongest baseline, Freezing and Denoising Graph Structures for Multimodal Recommendation (FREEDOM), HGCRec improves Recall@20 by 9.65%, 10.44%, and 12.00% on Baby, Sports, and Clothing, respectively, and improves NDCG@20 by 11.14%, 11.35%, and 13.10%. Sparsity-aware analysis further shows larger relative improvements for items with extremely limited training interactions, demonstrating the effectiveness of HGCRec under the evaluated interaction-sparse and few-shot settings.
- Open Access
- 2 August 2026
- Published by MDPI AG in Electronics
- Vol. 15 (15) , 3414
Collaborative Multimodal Entity Linking via Multi-Channel Neural Cross-Modal Interaction with Synergistic Consistency Optimization
Multimodal entity linking (MEL) grounds entity mentions in text-image contexts to entries in a structured knowledge base; however, most existing systems still decompose a multimodal document into independent mention-level decisions. This formulation overlooks a central tension of real-world MEL: the evidence needed to disambiguate an ambiguous mention is often distributed across co-occurring mentions, visual context, and cross-modal consistency, rather than being contained in the mention itself. To address this limitation, we propose Collaborative Entity Linking through Multichannel Interaction (CELMI), a four-channel framework that jointly models textual semantics, visual perception, cross-modal alignment, and inter-mention collaboration. CELMI employs dual-level textual alignment, text-guided visual gating, learned cross-modal projection, and attention-based entity graph propagation. To stabilize joint optimization, we further introduce a multi-channel consistency objective that combines per-channel contrastive losses with an overall ranking loss, reducing channel dominance and representation collapse. Among conventional non-LLM/VLM MEL models, CELMI achieves the strongest MRR and Hits@1 performance on WikiMEL, RichpediaMEL, and WikiDiverse, with Hits@1 scores of 89.03% on WikiMEL and 83.02% on RichpediaMEL; LLM/VLM systems remain stronger on WikiDiverse, positioning CELMI as a lightweight complement. Progressive stress ablation shows a 38.50 percentage-point absolute Hits@1 drop when the architecture is reduced to a single unimodal endpoint, confirming that the gains arise from synergistic channel interaction rather than isolated module effects.
- 2 August 2026
- Published by MDPI AG in Electronics
- Vol. 15 (15) , 3414
Open AccessCollaborative Multimodal Entity Linking via Multi-Channel Neural Cross-Modal Interaction with Synergistic Consistency Optimization
Multimodal entity linking (MEL) grounds entity mentions in text-image contexts to entries in a structured knowledge base; however, most existing systems still decompose a multimodal document into independent mention-level decisions. This formulation overlooks a central tension of real-world MEL: the evidence needed to disambiguate an ambiguous mention is often distributed across co-occurring mentions, visual context, and cross-modal consistency, rather than being contained in the mention itself. To address this limitation, we propose Collaborative Entity Linking through Multichannel Interaction (CELMI), a four-channel framework that jointly models textual semantics, visual perception, cross-modal alignment, and inter-mention collaboration. CELMI employs dual-level textual alignment, text-guided visual gating, learned cross-modal projection, and attention-based entity graph propagation. To stabilize joint optimization, we further introduce a multi-channel consistency objective that combines per-channel contrastive losses with an overall ranking loss, reducing channel dominance and representation collapse. Among conventional non-LLM/VLM MEL models, CELMI achieves the strongest MRR and Hits@1 performance on WikiMEL, RichpediaMEL, and WikiDiverse, with Hits@1 scores of 89.03% on WikiMEL and 83.02% on RichpediaMEL; LLM/VLM systems remain stronger on WikiDiverse, positioning CELMI as a lightweight complement. Progressive stress ablation shows a 38.50 percentage-point absolute Hits@1 drop when the architecture is reduced to a single unimodal endpoint, confirming that the gains arise from synergistic channel interaction rather than isolated module effects.
- Open Access
A Data-Efficient Surrogate Model via Simplified Feature Extraction and Pre-Training for Automatic History Matching
Automatic history matching traditionally depends on a large number of time-consuming numerical simulations, which makes the overall workflow computationally expensive. Deep learning-based surrogate models provide an efficient alternative, but their predictive performance often relies on large labeled datasets, whose generation through reservoir simulation remains costly. To alleviate this issue, we propose a data-efficient surrogate modeling framework for automatic history matching. The framework consists of two components. First, the reservoir parameter field is reformulated as a flattened representation and processed using one-dimensional convolutions. This representation provides a direct connection between parameter encoding and production-sequence prediction while maintaining competitive forecasting accuracy. Second, an autoencoder is pre-trained on unlabeled parameter realizations, and the learned encoder is then used to initialize the surrogate model for supervised regression, thereby improving the utilization of inexpensive, unlabeled data. The proposed framework is evaluated on the three-dimensional Brugge benchmark reservoir model. Results show that the one-dimensional representation achieves competitive predictive accuracy with shorter training time. In addition, the pre-training strategy is particularly beneficial when labeled simulation data are limited. Overall, the proposed framework improves the data efficiency of surrogate-assisted automatic history matching and reduces the dependence on extensive labeled simulations in the Brugge benchmark.
Open AccessA Data-Efficient Surrogate Model via Simplified Feature Extraction and Pre-Training for Automatic History Matching
Automatic history matching traditionally depends on a large number of time-consuming numerical simulations, which makes the overall workflow computationally expensive. Deep learning-based surrogate models provide an efficient alternative, but their predictive performance often relies on large labeled datasets, whose generation through reservoir simulation remains costly. To alleviate this issue, we propose a data-efficient surrogate modeling framework for automatic history matching. The framework consists of two components. First, the reservoir parameter field is reformulated as a flattened representation and processed using one-dimensional convolutions. This representation provides a direct connection between parameter encoding and production-sequence prediction while maintaining competitive forecasting accuracy. Second, an autoencoder is pre-trained on unlabeled parameter realizations, and the learned encoder is then used to initialize the surrogate model for supervised regression, thereby improving the utilization of inexpensive, unlabeled data. The proposed framework is evaluated on the three-dimensional Brugge benchmark reservoir model. Results show that the one-dimensional representation achieves competitive predictive accuracy with shorter training time. In addition, the pre-training strategy is particularly beneficial when labeled simulation data are limited. Overall, the proposed framework improves the data efficiency of surrogate-assisted automatic history matching and reduces the dependence on extensive labeled simulations in the Brugge benchmark.
- Open Access
- 10 October 2025
- Published by Springer Nature in Complex & Intelligent Systems
- Vol. 11 (11) , 1-19
A multi-modal entity alignment method based on image enhancement and credibility iteration
- 10 October 2025
- Published by Springer Nature in Complex & Intelligent Systems
- Vol. 11 (11) , 1-19
Open AccessA multi-modal entity alignment method based on image enhancement and credibility iteration
- Open Access
- 4 September 2025
- Published by MDPI AG in Information
- Vol. 16 (9) , 768
Secure Communication and Dynamic Formation Control of Intelligent Drone Swarms Using Blockchain Technology
With the increasing deployment of unmanned aerial vehicle (UAV) swarms in scenarios such as disaster response, environmental monitoring, and military reconnaissance, the need for secure and scalable formation control has become critical. Traditional centralized architectures face challenges such as limited scalability, communication bottlenecks, and single points of failure in large-scale swarm coordination. To address these issues, this paper proposes a blockchain-based decentralized formation control framework that integrates smart contracts to manage UAV registration, identity authentication, formation assignment, and positional coordination. The system follows a leader–follower structure, where the leader broadcasts formation tasks via on-chain events, while followers respond in real-time through event-driven mechanisms. A parameterized control model based on dynamic angle and distance adjustments is employed to support various formations, including V-shape, line, and circular configurations. The transformation from relative to geographic positions is achieved using Haversine and Euclidean methods. Experimental validation in a simulated environment demonstrates that the proposed method achieves lower communication latency and better responsiveness compared to polling-based schemes, while offering enhanced scalability and robustness. This work provides a feasible and secure decentralized control solution for future UAV swarm systems.
- 4 September 2025
- Published by MDPI AG in Information
- Vol. 16 (9) , 768
Open AccessSecure Communication and Dynamic Formation Control of Intelligent Drone Swarms Using Blockchain Technology
With the increasing deployment of unmanned aerial vehicle (UAV) swarms in scenarios such as disaster response, environmental monitoring, and military reconnaissance, the need for secure and scalable formation control has become critical. Traditional centralized architectures face challenges such as limited scalability, communication bottlenecks, and single points of failure in large-scale swarm coordination. To address these issues, this paper proposes a blockchain-based decentralized formation control framework that integrates smart contracts to manage UAV registration, identity authentication, formation assignment, and positional coordination. The system follows a leader–follower structure, where the leader broadcasts formation tasks via on-chain events, while followers respond in real-time through event-driven mechanisms. A parameterized control model based on dynamic angle and distance adjustments is employed to support various formations, including V-shape, line, and circular configurations. The transformation from relative to geographic positions is achieved using Haversine and Euclidean methods. Experimental validation in a simulated environment demonstrates that the proposed method achieves lower communication latency and better responsiveness compared to polling-based schemes, while offering enhanced scalability and robustness. This work provides a feasible and secure decentralized control solution for future UAV swarm systems.
- 30 June 2025
- Published by World Scientific Pub Co Pte Ltd in International Journal of Software Engineering and Knowledge Engineering
- Vol. 35 (07) , 943-961
Advances and Challenges in Multimodal Entity Linking: A Comprehensive Survey
This study offers a comprehensive examination of the recent advances and challenges in multimodal entity linking (MEL). The paper begins by outlining the background of the growth of multimodal data and the significance of knowledge bases. It provides a detailed description of the definition and architecture of the MEL task. Through an extensive analysis of existing methods, datasets, and evaluation metrics, this paper highlights the potential of MEL to enhance information retrieval accuracy and facilitate knowledge graph construction. The study indicates that, despite significant progress in the MEL field, challenges remain, including issues related to data quality, model generalization, model interpretability and multimodal fusion technologies. Finally, the paper proposes future research directions, such as the application of large language model and the development of scientific theories related to multimodal information fusion and disambiguation, thereby offering clear guidance for subsequent research.
- 30 June 2025
- Published by World Scientific Pub Co Pte Ltd in International Journal of Software Engineering and Knowledge Engineering
- Vol. 35 (07) , 943-961
Advances and Challenges in Multimodal Entity Linking: A Comprehensive Survey
This study offers a comprehensive examination of the recent advances and challenges in multimodal entity linking (MEL). The paper begins by outlining the background of the growth of multimodal data and the significance of knowledge bases. It provides a detailed description of the definition and architecture of the MEL task. Through an extensive analysis of existing methods, datasets, and evaluation metrics, this paper highlights the potential of MEL to enhance information retrieval accuracy and facilitate knowledge graph construction. The study indicates that, despite significant progress in the MEL field, challenges remain, including issues related to data quality, model generalization, model interpretability and multimodal fusion technologies. Finally, the paper proposes future research directions, such as the application of large language model and the development of scientific theories related to multimodal information fusion and disambiguation, thereby offering clear guidance for subsequent research.
- Open Access
Multi-Modal Entity Alignment Based on Enhanced Relationship Learning and Multi-Layer Feature Fusion
Entity alignment is a critical technique for integrating diverse knowledge graphs. Although existing methods have achieved impressive success in traditional entity alignment, they may struggle to handle the complexities arising from interactions and dependencies in multi-modal knowledge. In this paper, a novel multi-modal entity alignment model called ERMF is proposed, which leverages distinct modal characteristics of entities to identify equivalent entities across different multi-modal knowledge graphs. The symmetry in cross-modal interactions and hierarchical feature fusion is a core design principle of our approach. Specifically, we first utilize different feature encoders to independently extract features from different modalities. Concurrently, visual features and nearest neighbor negative sampling methods are incorporated to design a vision-guided negative sample generation strategy based on contrastive learning, ensuring a symmetric balance between positive and negative samples and guiding the model to learn effective relationship embeddings. Subsequently, in the feature fusion stage, we propose a multi-layer feature fusion approach that incorporates cross-attention and cross-modal attention mechanisms with symmetric processing of intra- and inter-modal correlations, thereby obtaining multi-granularity features. Extensive experiments were conducted on two public datasets, namely FB15K-DB15K and FB15K-YAGO15K. With 20% aligned seeds, ERMF improves Hits@1 by 8.4% and 26%, and MRR by 6% and 19.2% compared to the best baseline. The symmetric architecture of our model ensures the robust and balanced utilization of multi-modal information, aligning with the principles of structural and functional symmetry in knowledge integration.
Open AccessMulti-Modal Entity Alignment Based on Enhanced Relationship Learning and Multi-Layer Feature Fusion
Entity alignment is a critical technique for integrating diverse knowledge graphs. Although existing methods have achieved impressive success in traditional entity alignment, they may struggle to handle the complexities arising from interactions and dependencies in multi-modal knowledge. In this paper, a novel multi-modal entity alignment model called ERMF is proposed, which leverages distinct modal characteristics of entities to identify equivalent entities across different multi-modal knowledge graphs. The symmetry in cross-modal interactions and hierarchical feature fusion is a core design principle of our approach. Specifically, we first utilize different feature encoders to independently extract features from different modalities. Concurrently, visual features and nearest neighbor negative sampling methods are incorporated to design a vision-guided negative sample generation strategy based on contrastive learning, ensuring a symmetric balance between positive and negative samples and guiding the model to learn effective relationship embeddings. Subsequently, in the feature fusion stage, we propose a multi-layer feature fusion approach that incorporates cross-attention and cross-modal attention mechanisms with symmetric processing of intra- and inter-modal correlations, thereby obtaining multi-granularity features. Extensive experiments were conducted on two public datasets, namely FB15K-DB15K and FB15K-YAGO15K. With 20% aligned seeds, ERMF improves Hits@1 by 8.4% and 26%, and MRR by 6% and 19.2% compared to the best baseline. The symmetric architecture of our model ensures the robust and balanced utilization of multi-modal information, aligning with the principles of structural and functional symmetry in knowledge integration.
- 10 March 2025
- Published by Springer Nature in Archives of Computational Methods in Engineering
- Vol. 32 (7) , 4125-4162
A Review of Parallel Computing for Large-scale Reservoir Numerical Simulation
- 10 March 2025
- Published by Springer Nature in Archives of Computational Methods in Engineering
- Vol. 32 (7) , 4125-4162
A Review of Parallel Computing for Large-scale Reservoir Numerical Simulation
- Open Access
- 18 October 2024
- Published by Springer Nature in Research Square
Multi-modal Entity Alignment based on Enhanced Relationship Learning and Multi-layer Feature Fusion
Entity alignment is a critical technique for integrating diverse knowledge graphs. Although existing methods have achieved impressive success in traditional entity alignment, they may struggle to handle the complexities arising from interactions and dependencies in multi-modal knowledge. In this paper, we propose a novel multi-modal entity alignment model called ERMF, which leverages distinct modal characteristics of entities to identify equivalent entities across different multi-modal knowledge graphs. Specifically, we first utilize different feature encoders to independently extract features from different modalities. Concurrently, we incorporate visual features and random sampling methods to design a vision-guided negative sample generation strategy based on contrastive learning, guiding the model to learn relationship embeddings. Subsequently, in the feature fusion stage, we propose a multi-layer feature fusion approach that incorporates multiple attention mechanisms to hierarchically process the importance weights and interactions of the modalities, thereby obtaining multi-granularity features of the modalities. Extensive experiments were conducted on two public datasets, and the results demonstrated that ERMF significantly outperforms competitive baseline models, confirming the effectiveness of the proposed model.
- 18 October 2024
- Published by Springer Nature in Research Square
Open AccessMulti-modal Entity Alignment based on Enhanced Relationship Learning and Multi-layer Feature Fusion
Entity alignment is a critical technique for integrating diverse knowledge graphs. Although existing methods have achieved impressive success in traditional entity alignment, they may struggle to handle the complexities arising from interactions and dependencies in multi-modal knowledge. In this paper, we propose a novel multi-modal entity alignment model called ERMF, which leverages distinct modal characteristics of entities to identify equivalent entities across different multi-modal knowledge graphs. Specifically, we first utilize different feature encoders to independently extract features from different modalities. Concurrently, we incorporate visual features and random sampling methods to design a vision-guided negative sample generation strategy based on contrastive learning, guiding the model to learn relationship embeddings. Subsequently, in the feature fusion stage, we propose a multi-layer feature fusion approach that incorporates multiple attention mechanisms to hierarchically process the importance weights and interactions of the modalities, thereby obtaining multi-granularity features of the modalities. Extensive experiments were conducted on two public datasets, and the results demonstrated that ERMF significantly outperforms competitive baseline models, confirming the effectiveness of the proposed model.