A5076619169- Privacy-Preserving Technologies in Data
- Indoor and Outdoor Localization Technologies
- IoT and Edge/Fog Computing
- Internet Traffic Analysis and Secure E-voting
- Wireless Networks and Protocols
- Distributed Sensor Networks and Detection Algorithms
- Radar Systems and Signal Processing
- Context-Aware Activity Recognition Systems
- IoT Networks and Protocols
- Optical Systems and Laser Technology
- Lung Cancer Diagnosis and Treatment
- Vehicular Ad Hoc Networks (VANETs)
- Radiomics and Machine Learning in Medical Imaging
- Natural Language Processing Techniques
- Age of Information Optimization
- Mobile Crowdsensing and Crowdsourcing
- Domain Adaptation and Few-Shot Learning
- Human Mobility and Location-Based Analysis
- PAPR reduction in OFDM
- AI in cancer detection
- Antenna Design and Optimization
- Advanced SAR Imaging Techniques
- Topic Modeling
- Cooperative Communication and Network Coding
Beijing University of Posts and Telecommunications
2022-2025
Duke Kunshan University
2024
As a popular distributed learning paradigm, federated (FL) over mobile devices fosters numerous applications, while their practical deployment is hindered by participating devices' computing and communication heterogeneity. Some pioneering research efforts proposed to extract subnetworks from the global model, assign as large subnetwork possible device for local training based on its full capacity. Although such fixed size assignment enables FL heterogeneous devices, it unaware of (i)...
Training latency is critical for the success of numerous intrigued applications ignited by federated learning (FL) over heterogeneous mobile devices. By revolutionarily overlapping local gradient transmission with continuous computing, FL can remarkably reduce its training homogeneous clients, yet encounter severe model staleness, drifts, memory cost and straggler issues in environments. To unleash full potential overlapping, we propose, FedEx, a novel \underline{fed}erated approach to...
Cross-domain Relation Extraction aims to transfer knowledge from a source domain different target address low-resource challenges. However, the semantic gap caused by data bias between domains is major challenge, especially in few-shot scenarios. Previous work has mainly focused on transferring through shared feature representations without analyzing impact of each factor that may produce based characteristics domain. This takes causal perspective and proposes new framework CausalGF. By...
Participant selection (PS) helps to accelerate federated learning (FL) convergence, which is essential for the practical deployment of FL over mobile devices. While most existing PS approaches focus on improving training accuracy and efficiency rather than residual energy devices, fundamentally determines whether selected devices can participate. Meanwhile, impacts heterogeneous wireless transmission rates are largely ignored. Moreover, causes staleness issue. Prior research exploits...
Federated Learning (FL) algorithms commonly sample a random subset of clients to address the straggler issue and improve communication efficiency. While recent works have proposed various client sampling methods, they limitations in joint system data heterogeneity design, which may not align with practical heterogeneous wireless networks. In this work, we advocate new independent strategy minimize wall-clock training time FL, while considering both computation. We first derive convergence...
As a popular distributed learning paradigm, federated (FL) over mobile devices fosters numerous applications, while their practical deployment is hindered by participating devices' computing and communication heterogeneity. Some pioneering research efforts proposed to extract subnetworks from the global model, assign as large subnetwork possible device for local training based on its full communications capacity. Although such fixed size assignment enables FL heterogeneous devices, it...
Biomedical semantic segmentation aims to automat-ically label each pixel of a medical image with corresponding class, such as lung and heart. Such technique significantly facilitates clinical diagnosis thus has attracted increasing attentions both in industry academia. Although promising, existing methods U-Net MSRF, may suffer from the following two issues when they are applied ultrasound images: 1) few datasets publicly available due regulation patient's privacy, while U-NET/MSRF models...