A5114060150- Privacy-Preserving Technologies in Data
- Stochastic Gradient Optimization Techniques
- Brain Tumor Detection and Classification
- Anomaly Detection Techniques and Applications
- Advanced machining processes and optimization
- Tribology and Lubrication Engineering
- Advanced Data Storage Technologies
- Advanced Machining and Optimization Techniques
- Control Systems in Engineering
- Software System Performance and Reliability
- Traffic Prediction and Management Techniques
- Ferroelectric and Negative Capacitance Devices
- Video Surveillance and Tracking Methods
- IoT and Edge/Fog Computing
- Traffic and Road Safety
- Magnetic Bearings and Levitation Dynamics
- Internet Traffic Analysis and Secure E-voting
- Autonomous Vehicle Technology and Safety
- Advanced Surface Polishing Techniques
Changchun University of Science and Technology
2009-2024
Chongqing University of Science and Technology
2024
University of Macau
2024
This paper presents ProFL, a novel progressive FL framework to effectively break the memory wall. Specifically, ProFL divides model into different blocks based on its original architecture. Instead of updating full in each training round, first trains front and safely freezes them after convergence. Training next block is then triggered. process iterates until whole completed. In this way, footprint reduced for feasible deployment heterogeneous devices. order preserve feature representation...
This paper presents the techniques about microhole abrasive flow machining of crafts through controlling relative motion between workpiece and by speed cylinder piston, improving accuracy efficiency. Abrasive maching motor-driven piston is achieved using computer's information processing ability, changing pulse frequency programm based on Delphi language, direction stepping motor.
The autonomous displacement and measurement functions of the maglev ruler are performed by mover core. magnetic levitation can serve as a viable alternative to linear system coordinate measuring machine. stability four fields in air gaps position is one key factors for thrust force on power core, it also ensuring precision ruler. There cross-coupling between two ends core ruler, resulting strongly coupled, nonlinear, multi-input multi-output levitating This paper establishes mathematical...
Federated Learning (FL) emerges as a new learning paradigm that enables multiple devices to collaboratively train shared model while preserving data privacy. However, intensive memory footprint during the training process severely bottlenecks deployment of FL on resource-constrained in real-world cases. In this paper, we propose NeuLite, framework breaks wall through elastic progressive training. Unlike traditional FL, which updates full whole procedure, NeuLite divides into blocks and...
Federated Learning (FL) emerges as a new learning paradigm that enables multiple devices to collaboratively train shared model while preserving data privacy. However, one fundamental and prevailing challenge hinders the deployment of FL on mobile is memory limitation. This paper proposes FedHybrid, novel framework effectively reduces footprint during training process guaranteeing accuracy overall progress. Specifically, FedHybrid first selects participating for each round by jointly...
On-device training has become an increasingly popular approach to machine learning, enabling models be trained directly on mobile and edge devices. However, a major challenge in this area is the limited memory available these devices, which can severely restrict size complexity of that trained. In systematic survey, we aim explore current state-of-the-art techniques for breaking on-device walls, focusing methods enable larger more complex resource-constrained Specifically, first analyze key...