Yuan Ai
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Yuan Ai has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
Jia-Long Xu
Hainan Power Grid Co., Ltd, China Southern Power Grid, Haikou, China
2 shared publicationsLei Zhong
Hainan Power Grid Co., Ltd, China Southern Power Grid, Haikou, China
2 shared publicationsXue-Jiao Jiang
Hainan Power Grid Co., Ltd, China Southern Power Grid, Haikou, China
2 shared publications- Kaihong Zheng
China Southern Power Grid Digital Grid Group Company Ltd., Guangzhou, China
1 shared publication Lei Gao
Hainan Power Grid Co., Ltd, China Southern Power Grid, Haikou, China
1 shared publication- Lu-Kun Zeng
China Southern Power Grid Digital Grid Group Company Ltd., Guangzhou, China
1 shared publication
Latest Publications
- Open Access
- 19 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. PP, 1
Federated Contrastive Multi-hop Line Graph Attention Network for Link Prediction
Accurate link prediction is essential for optimizing energy consumption and operational efficiency, yet its implementation faces challenges due to private communication link and device data in certain regions of the distributed system—a scenario where federated learning emerges as a viable solution. While Line graph-based methods excel in link prediction of the centralized system by framing it as a node classification problem, they fall short in power system wireless networks where link existence hinges on both local node neighborhoods and global link conflicts. This limitation is further compounded by regional variations in distributed network topology, which cause distinct link patterns across areas. As a result, local models trained in different regions may develop divergent parameters, leading to skewed global model aggregation that overfits to specific areas and ultimately degrades link prediction performance. To address these limitations, this study proposes Federated Multi-hop Line Graph Attention Network with Contrastive Augmentation (FedMGAC), a novel link prediction framework that enhances both local and global modeling capabilities. The proposed approach constructs multi-hop subgraphs for target nodes during subgraph generation, which are then converted into line graphs to transform the graph classification problem into graph node classification, thereby mitigating information averaging problems. Additionally, a contrastive multi-scale graph neural network framework is established to emphasize the discriminative structural features of these multi-hop subgraphs. By leveraging a contrastive learning mechanism that creates positive and negative instances within the topological network, this framework captures the multi-scale structural features of links, enabling more accurate link prediction. In terms of global model aggregation, FedMGAC is designed with a regularization mechanism based on global model parameter constraints and a global aggregation strategy weighted by model parameters to prevent excessive disparities among client parameters. Experimental results demonstrate that the proposed FedMGAC algorithm outperforms traditional methods, validating the effectiveness of the multi-hop subgraphs, contrastive graph attention framework, and the global aggregation strategy.
- 19 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. PP, 1
Open AccessFederated Contrastive Multi-hop Line Graph Attention Network for Link Prediction
Accurate link prediction is essential for optimizing energy consumption and operational efficiency, yet its implementation faces challenges due to private communication link and device data in certain regions of the distributed system—a scenario where federated learning emerges as a viable solution. While Line graph-based methods excel in link prediction of the centralized system by framing it as a node classification problem, they fall short in power system wireless networks where link existence hinges on both local node neighborhoods and global link conflicts. This limitation is further compounded by regional variations in distributed network topology, which cause distinct link patterns across areas. As a result, local models trained in different regions may develop divergent parameters, leading to skewed global model aggregation that overfits to specific areas and ultimately degrades link prediction performance. To address these limitations, this study proposes Federated Multi-hop Line Graph Attention Network with Contrastive Augmentation (FedMGAC), a novel link prediction framework that enhances both local and global modeling capabilities. The proposed approach constructs multi-hop subgraphs for target nodes during subgraph generation, which are then converted into line graphs to transform the graph classification problem into graph node classification, thereby mitigating information averaging problems. Additionally, a contrastive multi-scale graph neural network framework is established to emphasize the discriminative structural features of these multi-hop subgraphs. By leveraging a contrastive learning mechanism that creates positive and negative instances within the topological network, this framework captures the multi-scale structural features of links, enabling more accurate link prediction. In terms of global model aggregation, FedMGAC is designed with a regularization mechanism based on global model parameter constraints and a global aggregation strategy weighted by model parameters to prevent excessive disparities among client parameters. Experimental results demonstrate that the proposed FedMGAC algorithm outperforms traditional methods, validating the effectiveness of the multi-hop subgraphs, contrastive graph attention framework, and the global aggregation strategy.
- Open Access
- 14 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 5809-5820
Weighted Correlation Attention Network-Based Dual-Phase Federated Learning Framework for Image Recognition
The recognition task of privacy-sensitive images requires effective federated learning (FL) models. However, there are challenges such as inconsistent data distributions, as well as the impact of different class correlations and class imbalance for classification across regions. Existing research on explicitly addressing class correlation and class imbalance within FL frameworks remains limited. Additionally, due to the heterogeneity and class imbalance across different clients, the global model’s parameter aggregation process faces issues of inconsistency, resulting in a significant discrepancy between the global model parameters and parts of local models’ parameters, which leads to poor classification performance on these clients’ local models. To address these challenges, we propose a novel dual-phase federated learning framework, Fd-WCAT, which incorporates a weighted correlation attention network for multi-label electricity image recognition. In this framework, each client trains a local model, constructing masked label correlation graphs. Then, the extracted label correlation embedding features are extracted from the graphs for a class imbalance-weighted classifier. To tackle the parameter inconsistency between local and global models during training, Fd-WCAT designs a dual-phase weighted loss function based on global-local parameter regularization. Each client computes its class imbalance coefficient and sends the local model parameters to the server. On the server side, local models are clustered into T groups to ensure similarity in model parameters within each group. Intra-group aggregation is performed to generate T temporary models with minimal inconsistencies. Then, imbalance weights for the prototype models are calculated based on the average imbalance coefficient of each group, and a global model is generated through imbalance weighted aggregation. Experimental results on multi-label benchmarks show that Fd-WCAT outperforms existing baseline models.
- 14 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 5809-5820
Open AccessWeighted Correlation Attention Network-Based Dual-Phase Federated Learning Framework for Image Recognition
The recognition task of privacy-sensitive images requires effective federated learning (FL) models. However, there are challenges such as inconsistent data distributions, as well as the impact of different class correlations and class imbalance for classification across regions. Existing research on explicitly addressing class correlation and class imbalance within FL frameworks remains limited. Additionally, due to the heterogeneity and class imbalance across different clients, the global model’s parameter aggregation process faces issues of inconsistency, resulting in a significant discrepancy between the global model parameters and parts of local models’ parameters, which leads to poor classification performance on these clients’ local models. To address these challenges, we propose a novel dual-phase federated learning framework, Fd-WCAT, which incorporates a weighted correlation attention network for multi-label electricity image recognition. In this framework, each client trains a local model, constructing masked label correlation graphs. Then, the extracted label correlation embedding features are extracted from the graphs for a class imbalance-weighted classifier. To tackle the parameter inconsistency between local and global models during training, Fd-WCAT designs a dual-phase weighted loss function based on global-local parameter regularization. Each client computes its class imbalance coefficient and sends the local model parameters to the server. On the server side, local models are clustered into T groups to ensure similarity in model parameters within each group. Intra-group aggregation is performed to generate T temporary models with minimal inconsistencies. Then, imbalance weights for the prototype models are calculated based on the average imbalance coefficient of each group, and a global model is generated through imbalance weighted aggregation. Experimental results on multi-label benchmarks show that Fd-WCAT outperforms existing baseline models.