Yiheng Sun
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yiheng Sun has a total of 17 co-authors. Here are the top co-authors by number of shared publications:
- Yao Zhang
Shanghai Key Laboratory of Data Science, School of Computer Science, Fudan University, Shanghai, China
3 shared publications - Yun Xiong
Fudan University, Shanghai, China
3 shared publications - Siwei Zhang
Fudan University, Shanghai, China
2 shared publications - Yangyong Zhu
College of Computer Science and Artificial Intelligence, Fudan Unviersity, Shanghai, China and Shanghai Data Research Institute, Shanghai, China
2 shared publications - Hongquan Liu
College of Computer Science and Artificial Intelligence, Fudan University, Shanghai, China
1 shared publication - Jiawei Zhang
University of California 95616 Davis CA USA
1 shared publication
Latest Publications
- 27 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Artificial Intelligence
- Vol. 7 (6) , 3237-3247
FedFlex: Protecting Shared Features in Vertical Federated Learning via Differential Privacy
Vertical Federated Learning (VFL) is a privacypreserving distributed learning paradigm that enables multiple parties to collaboratively train a model by sharing features derived from their local datasets. However, even without direct data sharing, these shared features can inadvertently expose sensitive information, rendering them susceptible to adversarial attacks such as reconstruction and inference. This paper introduces FedFlex, a novel VFL framework designed to safeguard sensitive information within shared features. While many existing approaches employ differential privacy (DP) to protect features by adding noise, they often compromise either model accuracy or protection generality. FedFlex addresses these shortcomings through a two-step integration of VFL and DP. Initially, it achieves generic protection by adding a task-agnostic amount of noise. Subsequently, it adaptively adjusts the scale and distribution of the features to be shared in a trainable manner, thereby enhancing model accuracy under the added noise. Experimental results demonstrate that FedFlex is resilient to prevalent privacy attacks and offers a superior balance between privacy and accuracy compared to prior methods.
- 27 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Artificial Intelligence
- Vol. 7 (6) , 3237-3247
FedFlex: Protecting Shared Features in Vertical Federated Learning via Differential Privacy
Vertical Federated Learning (VFL) is a privacypreserving distributed learning paradigm that enables multiple parties to collaboratively train a model by sharing features derived from their local datasets. However, even without direct data sharing, these shared features can inadvertently expose sensitive information, rendering them susceptible to adversarial attacks such as reconstruction and inference. This paper introduces FedFlex, a novel VFL framework designed to safeguard sensitive information within shared features. While many existing approaches employ differential privacy (DP) to protect features by adding noise, they often compromise either model accuracy or protection generality. FedFlex addresses these shortcomings through a two-step integration of VFL and DP. Initially, it achieves generic protection by adding a task-agnostic amount of noise. Subsequently, it adaptively adjusts the scale and distribution of the features to be shared in a trainable manner, thereby enhancing model accuracy under the added noise. Experimental results demonstrate that FedFlex is resilient to prevalent privacy attacks and offers a superior balance between privacy and accuracy compared to prior methods.
- Open Access
- 21 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 3174-3183
RDGSL: Dynamic Graph Representation Learning with Structure Learning
Temporal Graph Networks (TGNs) have shown remarkable performance in learning representation for continuous-time dynamic graphs. However, real-world dynamic graphs typically contain diverse and intricate noise. Noise can significantly degrade the quality of representation generation, impeding the effectiveness of TGNs in downstream tasks. Though structure learning is widely applied to mitigate noise in static graphs, its adaptation to dynamic graph settings poses two significant challenges. i) Noise dynamics. Existing structure learning methods are ill-equipped to address the temporal aspect of noise, hampering their effectiveness in such dynamic and ever-changing noise patterns. ii) More severe noise. Noise may be introduced along with multiple interactions between two nodes, leading to the re-pollution of these nodes and consequently causing more severe noise compared to static graphs. In this paper, we present RDGSL, a representation learning method in continuous-time dynamic graphs. Meanwhile, we propose dynamic graph structure learning, a novel supervisory signal that empowers RDGSL with the ability to effectively combat noise in dynamic graphs. To address the noise dynamics issue, we introduce the Dynamic Graph Filter, where we innovatively propose a dynamic noise function that dynamically captures both current and historical noise, enabling us to assess the temporal aspect of noise and generate a denoised graph. We further propose the Temporal Embedding Learner to tackle the challenge of more severe noise, which utilizes an attention mechanism to selectively turn a blind eye to noisy edges and hence focus on normal edges, enhancing the expressiveness for representation generation that remains resilient to noise. Our method demonstrates robustness towards downstream tasks, resulting in up to 5.1% absolute AUC improvement in evolving classification versus the second-best baseline.
- 21 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 3174-3183
Open AccessRDGSL: Dynamic Graph Representation Learning with Structure Learning
Temporal Graph Networks (TGNs) have shown remarkable performance in learning representation for continuous-time dynamic graphs. However, real-world dynamic graphs typically contain diverse and intricate noise. Noise can significantly degrade the quality of representation generation, impeding the effectiveness of TGNs in downstream tasks. Though structure learning is widely applied to mitigate noise in static graphs, its adaptation to dynamic graph settings poses two significant challenges. i) Noise dynamics. Existing structure learning methods are ill-equipped to address the temporal aspect of noise, hampering their effectiveness in such dynamic and ever-changing noise patterns. ii) More severe noise. Noise may be introduced along with multiple interactions between two nodes, leading to the re-pollution of these nodes and consequently causing more severe noise compared to static graphs. In this paper, we present RDGSL, a representation learning method in continuous-time dynamic graphs. Meanwhile, we propose dynamic graph structure learning, a novel supervisory signal that empowers RDGSL with the ability to effectively combat noise in dynamic graphs. To address the noise dynamics issue, we introduce the Dynamic Graph Filter, where we innovatively propose a dynamic noise function that dynamically captures both current and historical noise, enabling us to assess the temporal aspect of noise and generate a denoised graph. We further propose the Temporal Embedding Learner to tackle the challenge of more severe noise, which utilizes an attention mechanism to selectively turn a blind eye to noisy edges and hence focus on normal edges, enhancing the expressiveness for representation generation that remains resilient to noise. Our method demonstrates robustness towards downstream tasks, resulting in up to 5.1% absolute AUC improvement in evolving classification versus the second-best baseline.
- Open Access
- 21 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 3216-3225
iLoRE: Dynamic Graph Representation with Instant Long-term Modeling and Re-occurrence Preservation
Continuous-time dynamic graph modeling is a crucial task for many real-world applications, such as financial risk management and fraud detection. Though existing dynamic graph modeling methods have achieved satisfactory results, they still suffer from three key limitations, hindering their scalability and further applicability. i) Indiscriminate updating. For incoming edges, existing methods would indiscriminately deal with them, which may lead to more time consumption and unexpected noisy information. ii) Ineffective node-wise long-term modeling. They heavily rely on recurrent neural networks (RNNs) as a backbone, which has been demonstrated to be incapable of fully capturing node-wise long-term dependencies in event sequences. iii) Neglect of re-occurrence patterns. Dynamic graphs involve the repeated occurrence of neighbors that indicates their importance, which is disappointedly neglected by existing methods. In this paper, we present iLoRE, a novel dynamic graph modeling method with instant node-wise Long-term modeling and Re-occurrence preservation. To overcome the indiscriminate updating issue, we introduce the Adaptive Short-term Updater module that will automatically discard the useless or noisy edges, ensuring iLoRE's effectiveness and instant ability. We further propose the Long-term Updater to realize more effective node-wise long-term modeling, where we innovatively propose the Identity Attention mechanism to empower a Transformer-based updater, bypassing the limited effectiveness of typical RNN-dominated designs. Finally, the crucial re-occurrence patterns are also encoded into a graph module for informative representation learning, which will further improve the expressiveness of our method. Our experimental results on real-world datasets demonstrate the effectiveness of our iLoRE for dynamic graph modeling.
- 21 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 3216-3225
Open AccessiLoRE: Dynamic Graph Representation with Instant Long-term Modeling and Re-occurrence Preservation
Continuous-time dynamic graph modeling is a crucial task for many real-world applications, such as financial risk management and fraud detection. Though existing dynamic graph modeling methods have achieved satisfactory results, they still suffer from three key limitations, hindering their scalability and further applicability. i) Indiscriminate updating. For incoming edges, existing methods would indiscriminately deal with them, which may lead to more time consumption and unexpected noisy information. ii) Ineffective node-wise long-term modeling. They heavily rely on recurrent neural networks (RNNs) as a backbone, which has been demonstrated to be incapable of fully capturing node-wise long-term dependencies in event sequences. iii) Neglect of re-occurrence patterns. Dynamic graphs involve the repeated occurrence of neighbors that indicates their importance, which is disappointedly neglected by existing methods. In this paper, we present iLoRE, a novel dynamic graph modeling method with instant node-wise Long-term modeling and Re-occurrence preservation. To overcome the indiscriminate updating issue, we introduce the Adaptive Short-term Updater module that will automatically discard the useless or noisy edges, ensuring iLoRE's effectiveness and instant ability. We further propose the Long-term Updater to realize more effective node-wise long-term modeling, where we innovatively propose the Identity Attention mechanism to empower a Transformer-based updater, bypassing the limited effectiveness of typical RNN-dominated designs. Finally, the crucial re-occurrence patterns are also encoded into a graph module for informative representation learning, which will further improve the expressiveness of our method. Our experimental results on real-world datasets demonstrate the effectiveness of our iLoRE for dynamic graph modeling.
- Open Access
- 30 April 2023
- Published by Association for Computing Machinery (ACM)
- p. 478-488
TIGER: Temporal Interaction Graph Embedding with Restarts
Temporal interaction graphs (TIGs), consisting of sequences of timestamped interaction events, are prevalent in fields like e-commerce and social networks. To better learn dynamic node embeddings that vary over time, researchers have proposed a series of temporal graph neural networks for TIGs. However, due to the entangled temporal and structural dependencies, existing methods have to process the sequence of events chronologically and consecutively to ensure node representations are up-to-date. This prevents existing models from parallelization and reduces their flexibility in industrial applications. To tackle the above challenge, in this paper, we propose TIGER, a TIG embedding model that can restart at any timestamp. We introduce a restarter module that generates surrogate representations acting as the warm initialization of node representations. By restarting from multiple timestamps simultaneously, we divide the sequence into multiple chunks and naturally enable the parallelization of the model. Moreover, in contrast to previous models that utilize a single memory unit, we introduce a dual memory module to better exploit neighborhood information and alleviate the staleness problem. Extensive experiments on four public datasets and one industrial dataset are conducted, and the results verify both the effectiveness and the efficiency of our work.
- 30 April 2023
- Published by Association for Computing Machinery (ACM)
- p. 478-488
Open AccessTIGER: Temporal Interaction Graph Embedding with Restarts
Temporal interaction graphs (TIGs), consisting of sequences of timestamped interaction events, are prevalent in fields like e-commerce and social networks. To better learn dynamic node embeddings that vary over time, researchers have proposed a series of temporal graph neural networks for TIGs. However, due to the entangled temporal and structural dependencies, existing methods have to process the sequence of events chronologically and consecutively to ensure node representations are up-to-date. This prevents existing models from parallelization and reduces their flexibility in industrial applications. To tackle the above challenge, in this paper, we propose TIGER, a TIG embedding model that can restart at any timestamp. We introduce a restarter module that generates surrogate representations acting as the warm initialization of node representations. By restarting from multiple timestamps simultaneously, we divide the sequence into multiple chunks and naturally enable the parallelization of the model. Moreover, in contrast to previous models that utilize a single memory unit, we introduce a dual memory module to better exploit neighborhood information and alleviate the staleness problem. Extensive experiments on four public datasets and one industrial dataset are conducted, and the results verify both the effectiveness and the efficiency of our work.