A5043971547- Privacy-Preserving Technologies in Data
- Stochastic Gradient Optimization Techniques
- Reinforcement Learning in Robotics
- Advanced Neural Network Applications
- Evolutionary Game Theory and Cooperation
- Complex Network Analysis Techniques
- Bioinformatics and Genomic Networks
- Mobile Crowdsensing and Crowdsourcing
- Machine Learning and ELM
- Parallel Computing and Optimization Techniques
- Domain Adaptation and Few-Shot Learning
- Advanced Algorithms and Applications
- Power Systems and Renewable Energy
- Neural Networks and Applications
- Adversarial Robustness in Machine Learning
- Traffic Prediction and Management Techniques
- Modular Robots and Swarm Intelligence
- Network Security and Intrusion Detection
- Neural Networks and Reservoir Computing
- Advanced Decision-Making Techniques
- Human Mobility and Location-Based Analysis
- Adaptive Dynamic Programming Control
- Embedded Systems Design Techniques
- Advanced Graph Neural Networks
- Internet Traffic Analysis and Secure E-voting
Northwestern Polytechnical University
2018-2024
Vrije Universiteit Brussel
2022-2023
Lanzhou University
2017
In complex networks, the existing link prediction methods primarily focus on internal structural information derived from single-layer networks. However, role of interlayer is hardly recognized in multiplex which provide more diverse features than Actually, properties and functions one layer can affect that other layers this paper, effect performance investigated By utilizing intralayer information, we propose a novel “Node Similarity Index” based “Layer Relevance” (NSILR) network for...
Link prediction is to calculate the probability of a potential link between pair unlinked nodes in future. It has significance value both theoretical and practical. The similarity two networks an essential factor determine them. One important methods with consider common neighbors nodes. However, number only describes kind quantitative relationship without taking into account topology given information local structure which consist their neighbors. Therefore, we introduce concept degrees...
Formulae display:?Mathematical formulae have been encoded as MathML and are displayed in this HTML version using MathJax order to improve their display. Uncheck the box turn off. This feature requires Javascript. Click on a formula zoom.
Federated learning, as a distributed machine learning framework that shared global model is obtained through frequent local training parameter interaction on each participated device. However, the limited communication bandwidth of participating IoT and edge devices will have conflict between parameter-interaction mode federated impact efficiency. In this paper, efficiency enhanced technique presented by proposing cooperative filter selection method. The Geometric Median layer in adopted...
Federated learning (FL) on the edge devices must support continual (CL) to handle continuously evolving data and perform model training in an energy-efficient manner accommodate with limited computational energy resources. This letter proposes personalized federated CL (FCL) framework for devices. The network structure each device is divided into parts retaining old knowledge new knowledge, only part of reduce overhead. A data-free parameter selection approach selects important parameters...
The multi-objective neural architecture search (NAS) can automatically realize the network design for high accuracy and hardware performance varied applications. However, existing methods usually need to sample a large number of networks in process guide controller's behavior. As result, entire requires huge time overhead. We propose NAS framework that perform dynamic adaptive sampling regulated by latency requirements specific devices application scenarios. In layer-wise process,...
A photovoltaic generation system is affected by many factors. The traditional method of forecasting power more complicated. double-level neural network designed with consideration these factors: the first level calculates a real solar radiation input theory and weather coefficient; second final result which comes from highest temperature forecast. results indicate this has good forecast capacity, calculate are very close to measure value errors less than single network.
New generation airborne embedded system has deployed Graphical Processing Units (GPUs) to raise processing capability meet growing computational demands. Applications in the have strict real-time constraints. Therefore, it is necessary accurately predict timing behaviors of those applications. Many previous work propose GPU performance models estimate execution time However, most do not consider impact co-execution on performance. In this paper, we a workload-aware model applications...
Federated learning (FL), a new distributed technology, allows us to train the global model on edge and embedded devices without local data sharing. However, due wide distribution of different types devices, FL faces severe heterogeneity issues. The accuracy efficiency deployment at are severely impacted by heterogeneous systems. In this paper, we perform joint personalization for systems address challenges posed heterogeneities. We begin using inference as starting point personalize network...
Heterogeneities have emerged as a critical challenge in Federated Learning (FL). In this paper, we identify the cause of FL performance degradation due to heterogeneous issues: local communicated parameters feature mismatches and representation range mismatches, resulting ineffective global model generalization. To address it, Heterogeneous mitigating is proposed improve generalization with resource-independence aggregation. Instead linking contributions its occupied resources, look for...
Federated Learning (FL) is a popular method for privacy-preserving machine learning on edge devices. However, the heterogeneity of devices, including differences in system architecture, data, and co-running applications, can significantly impact energy efficiency FL. To address these issues, we propose an energy-efficient personalized federated search framework. This framework has three key components. Firstly, partial models with high inference to reduce training consumption occurrence...
Intelligent unmanned systems (IUSs) are distributed composed of multiple agents that share information or cooperate to accomplish specific complex tasks. Agents the IUS capable perception, cognition, control, decision-making, and action. In some cases, environmental situation task objectives faced by IUSs constantly changing with time. Thus, time-sensitive systems. To accelerate execution time response speed, use artificial intelligence technology increase speed quality...