A5100373563- Imbalanced Data Classification Techniques
- Advanced Computational Techniques and Applications
- Electricity Theft Detection Techniques
- Human Pose and Action Recognition
- Anomaly Detection Techniques and Applications
- Rough Sets and Fuzzy Logic
- Generative Adversarial Networks and Image Synthesis
- Service-Oriented Architecture and Web Services
- Video Surveillance and Tracking Methods
- Medical Image Segmentation Techniques
- Artificial Intelligence in Healthcare
- Music and Audio Processing
- Vehicle License Plate Recognition
- Cognitive Computing and Networks
- Facility Location and Emergency Management
- Voice and Speech Disorders
- Human Mobility and Location-Based Analysis
- Face recognition and analysis
- Advanced Neuroimaging Techniques and Applications
- Disaster Management and Resilience
- Advanced Image Processing Techniques
- Disaster Response and Management
- Data Mining Algorithms and Applications
- Domain Adaptation and Few-Shot Learning
- Neural Networks and Applications
University of Electronic Science and Technology of China
2019-2025
Chongqing University
2005-2024
Chongqing Academy of Environmental Science
2024
Louisiana Tech University
2024
Peking University First Hospital
2024
Peking University
2024
Chongqing Jiaotong University
2024
State Key Laboratory of Modern Optical Instruments
2023
Tongji University
2023
State Key Laboratory on Integrated Optoelectronics
2023
Real-world black-box optimization often involves time-consuming or costly experiments and simulations. Multi-fidelity (MFO) stands out as a cost-effective strategy that balances high-fidelity accuracy with computational efficiency through hierarchical fidelity approach. This survey presents systematic exploration of MFO, underpinned by novel text mining framework based on pre-trained language model. We delve deep into the foundational principles methodologies focusing three core components...
Medical insurance plays a vital role in modern society, yet organized healthcare fraud causes billions of dollars annual losses, severely harming the sustainability social welfare system. Existing works mostly focus on detecting individual entities or claims, ignoring hidden conspiracy patterns. Hence, they face severe challenges tackling fraud. In this paper, we proposed RDPGL, novel Risk Diffusion-based Parallel Graph Learning approach, to fighting against medical criminal gangs....
Oil-based drilling cuttings (OBDCs) contain petroleum hydrocarbons with complex compositions and high concentrations, which have highly carcinogenic, teratogenic, mutagenic properties. In this study, three efficient hydrocarbon-degrading bacteria were screened from OBDCs of different shale gas wells in Chongqing, China, identified as Rhodococcus sp. Dietzia Because their ability to degrade various chain lengths, a new method was proposed for degrading by combining bacterial species. Results...
Model selection methods based on stochastic regularization suchas Dropout have been widely used in deep learning due to theirsimplicity and effectiveness. The standard method treatsall units, visible or hidden, the same way, thus ignoring any a prioriinformation related grouping structure. Such structure ispresent multi-modal applications, where subsets of unitsmay correspond individual modalities. In this abstract we describeModout, model regularization,which is particularly useful...
Abstract AI-aided clinical diagnosis is desired in medical care. Existing deep learning models lack explainability and mainly focus on image analysis. The recently developed Dynamic Uncertain Causality Graph (DUCG) approach causality-driven, explainable, invariant across different application scenarios, without problems of data collection, labeling, fitting, privacy, bias, generalization, high cost energy consumption. Through close collaboration between experts DUCG technicians, 46 covering...
Using a computer to generate images with realistic is new direction in current vision research. This paper designs an image generation model based on the Generative Adversarial Network (GAN). creates - discriminator network and generator by eliminating fully connected layer traditional applying batch normalization deconvolution operations. also uses hyper-parameter measure diversity quality of generated image. The experimental results CelebA dataset show that has excellent performance face...
Amyloid β-protein (Aβ) plaque deposition is an important prevention and treatment target for Alzheimer's disease (AD). As a noninvasive, nonradioactive highly cost-effective clinical imaging method, magnetic resonance (MRI) the perfect technology diagnosis of AD, but it cannot display directly. This paper resolves this problem based on pixel feature selection algorithms at image level. Firstly, brain region was segmented from mouse model MR images. Secondly, pixels in were extracted as...
Traditional age estimation methods are based on the same idea that uses real as training label. However, these ignore there is a deviation between and brain due to accelerated aging.
Ensemble methods are widely used to tackle class imbalance problem. However, for existing imbalanced ensemble (IE) methods, the samples in each subset resampled from same dataset, and directly input classifier training, so quality (diversity separability) of subsets is unsatisfactory usually. To solve problem, a deep fuzzy envelope sample generation mechanism proposed. First, Fuzzy C-Means clustering based pre-network (DSEN) designed mine correlation information among samples, thereby...
Due to the high accuracy, favorable data share and large coverage, Automatic Dependent Surveillance - Broadcast (ADS-B) surveillance is regarded as core technology in next generation air traffic management. However, ADS-B broadcasted with absence of adequate integrity authentication support, which leads various security challenges on information leakage tampering system. Hence, anomaly detection vital minimize threats system application, especially for attack concealment. In this paper, a...
The rapid development of information and communication technology the popularization mobile devices have generated a large number spatiotemporal trajectory data. Trajectory data can be applied to location prediction, which is significant for urban traffic planning location-based service. Although various methods personal prediction been proposed, historical some users always sparse in practical applications, resulting poor precision models based on those users. Targeting this challenge, we...
An optical-electronic hybrid convolutional neural network (CNN) system is proposed and investigated for its parallel processing capability design robustness. It regarded as a practical way to implement real-time optical computing. In this paper, we propose complex-valued modulation method based on an amplitude-only liquid-crystal-on-silicon spatial light modulator fixed four-level diffractive element. A comparison of computational results convolutions between different methods in the Fourier...