A5048506332- Advanced Neural Network Applications
- Advanced Image and Video Retrieval Techniques
- Video Surveillance and Tracking Methods
- Domain Adaptation and Few-Shot Learning
- Multimodal Machine Learning Applications
- Human Pose and Action Recognition
- Hip disorders and treatments
- 3D Shape Modeling and Analysis
- Robotics and Sensor-Based Localization
- Orthopaedic implants and arthroplasty
- Stochastic Gradient Optimization Techniques
- Advanced Vision and Imaging
- Facial Rejuvenation and Surgery Techniques
- Ultrasound in Clinical Applications
- 3D Surveying and Cultural Heritage
- Anomaly Detection Techniques and Applications
- Osteoarthritis Treatment and Mechanisms
- Cloud Computing and Resource Management
- Adversarial Robustness in Machine Learning
- Cardiac Valve Diseases and Treatments
- Remote Sensing and LiDAR Applications
- Sparse and Compressive Sensing Techniques
- Cancer-related molecular mechanisms research
- Machine Learning and ELM
- Axon Guidance and Neuronal Signaling
Worcester Polytechnic Institute
2020-2025
Chinese Academy of Medical Sciences & Peking Union Medical College
2023-2025
Northeastern University
2006-2025
Tiangong University
2024
Hunan Academy of Traditional Chinese Medicine
2024
Hunan University of Traditional Chinese Medicine
2024
Nanjing University of Science and Technology
2024
Henan Agricultural University
2024
Union Hospital
2016-2024
Huazhong University of Science and Technology
2016-2024
Training a generic objectness measure to produce small set of candidate object windows, has been shown speed up the classical sliding window detection paradigm. We observe that objects with well-defined closed boundary can be discriminated by looking at norm gradients, suitable resizing their corresponding image windows in fixed size. Based on this observation and computational reasons, we propose resize 8 × use gradients as simple 64D feature describe it, for explicitly training measure....
In this paper we consider a version of the zero-shot learning problem where seen class source and target domain data are provided. The goal during test-time is to accurately predict label an unseen instance based on revealed side information (e.g. attributes) for classes. Our method viewing each or as mixture proportions postulate that patterns have be similar if two instances belong same class. This perspective leads us source/target embedding functions map arbitrary into semantic space...
Zero-shot recognition (ZSR) deals with the problem of predicting class labels for target domain instances based on source side information (e.g. attributes) unseen classes. We formulate ZSR as a binary prediction problem. Our resulting classifier is class-independent. It takes an arbitrary pair and input predicts whether or not they come from same class, i.e. there match. model posterior probability match since it sufficient statistic propose latent probabilistic in this context. develop...
Current methods for video description are based on encoder-decoder sentence generation using recurrent neural networks (RNNs). Recent work has demonstrated the advantages of integrating temporal attention mechanisms into these models, in which decoder network predicts each word by selectively giving more weight to encoded features from specific time frames. Such typically use two different types features: image (from an object classification model), and motion action recognition combined...
In this paper, we propose training very deep neural networks (DNNs) for supervised learning of hash codes. Existing methods in context train relatively "shallow" limited by the issues arising back propagation (e.g. vanishing gradients) as well computational efficiency. We a novel and efficient algorithm inspired alternating direction method multipliers (ADMM) that overcomes some these limitations. Our decomposes process into independent layer-wise local updates through auxiliary variables....
Abstract Background Bedside lung ultrasound (LUS) has emerged as a useful and non-invasive tool to detect involvement monitor changes in patients with coronavirus disease 2019 (COVID-19). However, the clinical significance of LUS score COVID-19 remains unknown. We aimed investigate prognostic value COVID-19. Method The protocol consisted 12 scanning zones was performed 280 consecutive based on B-lines, consolidation pleural line abnormalities evaluated. Results median time from admission...
In this paper we consider a version of the zero-shot learning problem where seen class source and target domain data are provided. The goal during test-time is to accurately predict label an unseen instance based on revealed side information (\eg attributes) for classes. Our method viewing each or as mixture proportions postulate that patterns have be similar if two instances belong same class. This perspective leads us source/target embedding functions map arbitrary into semantic space...
Object recognition has made great strides recently. However, the best methods, such as those based on kernel-SVMs are highly computationally intensive. The problem of how to accelerate evaluation process without decreasing accuracy is thus current interest. In this paper, we deal with by using idea ranking. We propose a cascaded architecture which ranking SVM generates an ordered set proposals for windows containing object instances. top may then be fed more complex detector. Our experiments...
The efficacy of anticancer drugs is often limited by their systemic toxicities and adverse side effects. We report that the EphA2 receptor overexpressed preferentially in several human cancer cell lines compared to normal tissues an targeting peptide (YSAYPDSVPMMS) can be effective delivering agents such tumors. Hence, we on synthesis characterizations a novel EphA2-targeting agent conjugated with chemotherapeutic drug paclitaxel. found peptide-drug conjugate dramatically more than...
In modern cloud computing systems, hundreds and even thousands of servers are interconnected by multi-layer networks. such large-scale complex failures common. Proactive failure management is a crucial technology to characterize system behaviors forecast dynamics in the cloud. To make predictions, we need monitor execution collect health-related runtime performance data. However, newly deployed or managed these data usually unlabeled. Supervised learning based approaches not suitable this...
Training a generic objectness measure to produce object proposals has recently become of significant interest. We observe that objects with well-defined closed boundaries can be detected by looking at the norm gradients, suitable resizing their corresponding image windows small fixed size. Based on this observation and computational reasons, we propose resize window 8 × use gradients as simple 64D feature describe it, for explicitly training measure. further show how binarized version...
In contrast to the literature where local patterns in 3D point clouds are captured by customized convolutional operators, this paper we study problem of how effectively and efficiently project such into a 2D image space so that traditional neural networks (CNNs) as U-Net can be applied for segmentation. To end, motivated graph drawing reformulate it an integer programming learn topology-preserving graph-to-grid mapping each individual cloud. accelerate computation practice, further propose...
Through screening by NMR spectroscopy, we discovered a novel scaffold (DPQ: 6,7-dimethoxy-2-(1-piperazinyl)-4-quinazolinamine) that binds specifically to the influenza A virus RNA promoter. The solution structure of RNA–DPQ complex reported here demonstrates internal loop is binding site DPQ. exhibits antiviral activity against viruses.
Abstract Purpose: YSA is an EphA2-targeting peptide that effectively delivers anticancer agents to prostate cancer tumors. Here, we report on how increased the drug-like properties of this delivery system. Experimental Design: By introducing non-natural amino acids, have designed two new EphA2 targeting peptides: YNH, where norleucine and homoserine replace methionine residues YSA, dYNH, a D-tyrosine replaces L-tyrosine at first position YNH peptide. We describe details synthesis dYNH...
To learn distinguishable patterns, most of recent works in vehicle re-identification (ReID) struggled to redevelop official benchmarks provide various supervisions, which requires prohibitive human labors. In this paper, we seek achieve the similar goal but do not involve more efforts. end, introduce a novel framework, successfully encodes both geometric local features and global representations distinguish instances, optimized only by supervision from ID labels. Specifically, given our...
Estimating homography to align image pairs captured by different sensors or with large appearance changes is an important and general challenge for many computer vision applications. In contrast others, we propose a generic solution pixel-wise multimodal extending the traditional Lucas-Kanade algorithm networks. The key contribution in our method how construct feature maps, named as deep map (DLKFM). learned DLKFM can spontaneously recognize invariant features under various...
Osteosarcoma (OS) mainly happens in children and youths. Surgery, radiotherapy chemotherapy are the common therapies for osteosarcoma treatment but all their anti-tumor effects limited. In recent years, a new cellular therapy, CAR-T, immunotherapy with genetically engineered T cells bearing chimeric antigen receptor targeting specific tumor-associated antigen, has been proved to be an effective therapy against acute lymphoblastic leukemia. Thus, CAR-T is potentially treatment.A CAR gene...
Quality improvement is crucial for manufacturing, and existing research has paid less attention to the influence of regulatory factors irrational decision makers. Considering impact reward punishment strategy shared platform on quality decision-making, this paper introduces prospect theory mental account into process multi-agent evolutionary game constructs a co-evolutionary model manufacturing synergistic under dynamic mechanism, analyzes evolution law each agent. The results show that: (1)...