A5032302838- Adversarial Robustness in Machine Learning
- Particle Detector Development and Performance
- Privacy-Preserving Technologies in Data
- Domain Adaptation and Few-Shot Learning
- Radiation Detection and Scintillator Technologies
- Medical Imaging Techniques and Applications
- Visual Attention and Saliency Detection
- Advanced Neural Network Applications
- Nuclear Physics and Applications
Shanghai University of Engineering Science
2023-2024
Institute of High Energy Physics
2022
The reconstruction of charged particle trajectories in tracking detectors is a key problem the analysis experimental data for high-energy and nuclear physics. amount modern experiments so large that classical methods, such as Kalman filter, cannot process them fast enough. To solve this problem, we developed two neural network algorithms based on deep learning architectures track recognition pixel strip-based detectors. These are TrackNETv3 local (track by track) RDGraphNet global (all...
When dealing with non-IID data, federated learning confronts issues such as client drift and sluggish convergence. Therefore, we propose a Bidirectional Corrective Model-Contrastive Federated Adversarial Training (BCMCFAT) framework. On the side, design category information correction module to correct biases caused by imbalanced local data incorporating client’s distribution information. Through adversarial training, more robust models are obtained. Secondly, model-based adaptive algorithm...
Salient object detection (SOD) networks are vulnerable to adversarial attacks. As training is computationally expensive for SOD, existing defense methods instead adopt a noise-against-noise strategy that disrupts perturbation and restores the image either in input or feature space. However, their limited learning capacity need network modifications limit applicability. In recent years, popular diffusion model coincides with idea exhibits excellent purification performance, but there still...
The reconstruction of charged particle trajectories in tracking detectors is a key problem the analysis experimental data for high-energy and nuclear physics. amount modern experiments so large that classical methods, such as Kalman filter, cannot process them fast enough. To solve this problem, we developed two neural network algorithms based on deep learning architectures track recognition pixel strip-based detectors. These are TrackNETv3 local (track by track) RDGraphNet global (all...