A5048331196- Advanced Neural Network Applications
- Radiomics and Machine Learning in Medical Imaging
- Machine Learning and ELM
- COVID-19 diagnosis using AI
- Face and Expression Recognition
- Medical Image Segmentation Techniques
- AI in cancer detection
- Imbalanced Data Classification Techniques
- Domain Adaptation and Few-Shot Learning
- Congenital Heart Disease Studies
- Brain Tumor Detection and Classification
- Advanced Measurement and Detection Methods
- Neural Networks and Applications
- Optical Systems and Laser Technology
- Phonocardiography and Auscultation Techniques
- Energy Load and Power Forecasting
- Electricity Theft Detection Techniques
- Infrared Target Detection Methodologies
- Traffic Prediction and Management Techniques
- Industrial Technology and Control Systems
- Autonomous Vehicle Technology and Safety
- Video Surveillance and Tracking Methods
- Privacy-Preserving Technologies in Data
- Cardiac Valve Diseases and Treatments
- Image Processing Techniques and Applications
Shenzhen University Health Science Center
2019-2025
Shenzhen University
2020-2024
Qingdao University of Science and Technology
2024
Hebei University of Environmental Engineering
2023
Capital Normal University
2021
Electric Power Research Institute
2020
University of Macau
2017-2019
ORCID
2019
Shijiazhuang University
2011-2012
Shandong Provincial Communications Planning and Design Institute (China)
2007
High accuracy of text classification can be achieved through simultaneous learning multiple information, such as sequence information and word importance. In this article, a kind flat neural networks called the broad system (BLS) is employed to derive two novel methods for classification, including recurrent BLS (R-BLS) long short-term memory (LSTM)-like architecture: gated (G-BLS). The proposed possess three advantages: 1) higher due even compared deep LSTM that extracts deeper but single...
Automatic gastric tumor segmentation and lymph node (LN) classification not only can assist radiologists in reading images, but also provide image-guided clinical diagnosis improve accuracy. However, due to the inhomogeneous intensity distribution of LN CT scans, ambiguous/missing boundaries, highly variable shapes tumor, it is quite challenging develop an automatic solution. To comprehensively address these challenges, we propose a novel 3D multi-attention guided multi-task learning network...
Predicting the motion trajectories of moving agents in complex traffic scenes, such as crossroads and roundabouts, plays an important role cooperative intelligent transportation systems. Nevertheless, accurately forecasting behavior a dynamic scenario is challenging due to interactions between agents. Graph Convolutional Neural Network has recently been employed deal with Despite promising performance resulting trajectory prediction algorithms, many existing graph-based approaches model...
Abstract With the development of big data and artificial intelligence, applications smart grid have received extensive attention. Specifically, accurate power system load forecasting plays an important role in safety stability production scheduling process. Due to limitations traditional methods dealing with large scale nonlinear time series data, this paper, we proposed Attention-BiLSTM (Attention based Bidirectional Long Short-Term Memory, Attention-BiLSTM) network do short-term...
Data heterogeneity across medical centers, resulting in a coupling of universal information for classification tasks and personalized private dataset within local models, is still difficult challenge Personalized Federated Learning (PFL). Moreover, the high inter-class similarity datasets affects performance models. Different from pervious works that focus on aggregation or adjusting global model, we introduce concept decoupling models propose novel PFL framework image this paper....
Current state-of-the-art class-imbalanced loss functions for deep models require exhaustive tuning on hyperparameters high model performance, resulting in low training efficiency and impracticality nonexpert users. To tackle this issue, a parameter-free (PF-loss) function is proposed, which works both binary multiclass-imbalanced learning image classification tasks. PF-loss provides three advantages: 1) time significantly reduced due to NO hyperparameter(s); 2) it dynamically pays more...
Automatic classification and segmentation of medical images play essential roles in computer-aided diagnosis. Deep convolutional neural networks (DCNNs) have shown their advantages on image segmentation. However, they not achieved the same success as done natural images. In this paper, two challenges are exploited for DCNNs images, including 1) lack feature diversity; 2) neglect small lesions. These issues heavily influence performances. To improve performance DCNN similarity-aware attention...
In this paper, a novel learning method called postboosting using extended G-mean (PBG) is proposed for online sequential multiclass imbalance (OS-MIL) in neural networks. PBG effective due to three reasons. 1) Through postadjusting classification boundary under G-mean, the challenging issue of imbalanced class distribution sequentially arriving data can be effectively resolved. 2) A newly derived update rule proposed, which produces high current model and simultaneously possesses almost same...
Accurate segmentation of pediatric echocardiography images is essential for a wide range diagnostic and pre-interventional planning, but remains challenging (e.g., low signal to noise ratio internal variability in heart appearance). To address these problems, this paper, we propose novel Cardiac Attention-guided Dual-path Network (i.e., AIDAN). AIDAN comprises convolutional block attention module (CBAM) attached spatial SPA) context paths CPA), which can guide the network learn most...
Deep metric learning (DML) has been widely applied in various tasks (e.g., medical diagnosis and face recognition) due to the effective extraction of discriminant features via reducing <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">data overlapping</i> . However, practice, these also easily suffer from two class-imbalance (CIL) problems: scarcity</i> density</i> , causing misclassification. Existing DML losses rarely consider issues, while...
In this paper, a robust online multilabel learning method dealing with dynamically changing data streams is proposed. The proposed has three advantages: 1) higher accuracy due to newly defined objective function based on labels ranking; 2) fast training and update derived closed-form (rather than gradient descent based) solution for the new function; 3) high robustness identified concept drift in streams, namely, changes distribution (CDDL). benefits from two novel works: sequential rule...
Machine learning aims to generate a predictive model from training dataset of fixed number known classes. However, many real-world applications (such as health monitoring and elderly care) are data streams in which new arrive continually short time. Such may even belong previously unknown Hence, class-incremental (CIL) is necessary, incrementally rapidly updates an existing with the classes while retaining knowledge old most current CIL methods designed based on deep models that require...
Pixel-level annotations are extremely expensive for medical image segmentation tasks as both expertise and time needed to generate accurate annotations. Semi-supervised learning (SSL) has recently attracted growing attention because it can alleviate the exhausting manual clinicians by leveraging unlabeled data. However, most of existing SSL methods do not take pixel-level information (e.g., features) labeled data into account, i.e., underutilized. Hence, in this work, an innovative <italic...