A5028872220- Adversarial Robustness in Machine Learning
- Anomaly Detection Techniques and Applications
- Machine Learning in Bioinformatics
- Advanced Image and Video Retrieval Techniques
- Network Security and Intrusion Detection
- SARS-CoV-2 and COVID-19 Research
- Image Retrieval and Classification Techniques
- Advanced Neural Network Applications
- COVID-19 diagnosis using AI
- vaccines and immunoinformatics approaches
- Domain Adaptation and Few-Shot Learning
- Advanced Malware Detection Techniques
- Parallel Computing and Optimization Techniques
- Graph Theory and Algorithms
- Distributed and Parallel Computing Systems
- Digital Media Forensic Detection
- Internet Traffic Analysis and Secure E-voting
- Advanced Graph Neural Networks
- Advanced Steganography and Watermarking Techniques
- Video Surveillance and Tracking Methods
- Acute Myocardial Infarction Research
- Nuclear Materials and Properties
- Knowledge Management and Technology
- Heart Rate Variability and Autonomic Control
- Generative Adversarial Networks and Image Synthesis
Yunnan University
2020-2025
National University of Singapore
2025
Nanyang Technological University
2024
Hubei University of Arts and Science
2023
Xiangyang Central Hospital
2023
Chinese Academy of Sciences
2021-2023
Wuhan University of Science and Technology
2023
Shenzhen Institutes of Advanced Technology
2023
Zhejiang Provincial People's Hospital
2022
Bridge University
2020-2021
Abstract Phosphate, a key plant nutrient, is perceived through inositol polyphosphates (InsPs) by SPX domain-containing proteins. SPX1 an inhibit the PHR2 transcription factor to maintain Pi homeostasis. How recognizes InsP molecule and represses activation remains unclear. Here we show that, upon binding 6 , can disrupt dimers form 1:1 SPX1-PHR2 complex. The complex structure reveals that helix α1 impose steric hindrance when interacting with dimer. By stabilizing α1, allosterically...
Existing black-box attacks have demonstrated promising potential in creating adversarial examples (AE) to deceive deep learning models. Most of these need handle a vast optimization space and require large number queries, hence exhibiting limited practical impacts real-world scenarios. In this paper, we propose novel attack strategy, Conditional Diffusion Model Attack (CDMA), improve the query efficiency generating AEs under query-limited situations. The key insight CDMA is formulate task AE...
Deep neural networks are vulnerable to adversarial attacks either by examples with indistinguishable perturbations which produce incorrect predictions, or noticeable transformations that still predicted as the original label. The latter case is known Type I attack which, however, has achieved limited attention in literature. We advocate vulnerability comes from ambiguous distributions among different classes resultant feature space of model, saying appearances may present similar features....
Abstract Copy–move forgery poses a significant threat to social life and has aroused much attention in recent years. Although many copy‐move detection (CMFD) methods have been proposed, the most existing CMFD are short of adaptability detecting images, which leads limitation on effects. To solve this problem, paper proposes novel keypoint‐based method: second‐keypoint matching double adaptive filtering (SMDAF). Motivated by image based keypoint, method is designed match keypoints extracted...
Recent adversarial attack research reveals the vulnerability of learning-based deep learning models (DNN) against well-designed perturbations. However, most existing methods have inherent limitations in image quality as they rely on a relatively loose noise budget, i.e., limit perturbations by L p -norm. Resulting that generated these can be easily detected defense mechanisms and are perceptible to human visual system (HVS). To circumvent former problem, we propose novel framework, called...
Extensive studies have demonstrated that deep neural networks (DNNs) are vulnerable to adversarial examples (AEs), which brings a huge security risk the application of DNNs, especially for AI models developed in real world. To impede process fully exploiting vulnerabilities existing DNNs and further improving their robustness face such malicious inputs, many attack methods been proposed build AEs. Despite significant progress has made recently, still suffer from unsatisfactory performance...
Background: This study assessed the effects of esmolol injection in patients with in-hospital cardiac arrest (IHCA) refractory ventricular fibrillation (VF)/pulseless tachycardia (pVT). Methods: From January 2018 to December 2021, 29 IHCA shockable rhythm were retrospectively reviewed. Esmolol was administered after advanced cardiovascular life support (ACLS)-directed procedures, and outcomes assessed. Results: Among cases, rates sustained return spontaneous circulation (ROSC), 24-h ROSC,...
Abstract The outbreak of novel coronavirus (SARS-CoV-2) developed into a global pandemic in few months. latest study found that the virus belongs to beta family. SARS-CoV-2 is highly similar Pangolin CoV and BatCoV RaTG. Advanced scientific studies help traceability vaccine development. In addition subgenus classification analysis virus, it interesting for further exploration focus attention on mutations their transmissions different regions. New may be likely affect symptoms disease...
Deep neural networks (DNNs) are sensitive to adversarial examples which generated by corrupting benign with imperceptible perturbations, or have significant changes but can still achieve original prediction results. The latter case is termed as the Type I example which, however, has limited attention in literature. In this paper, we introduce two methods, HRG and GAG, generate attempt apply them privacy-preserving Machine Learning a Service (MLaaS). Existing methods for MLaaS mostly based on...
Traditional black-box attack methods rely on sufficient feedback from the victim model through a large number of queries until is successful. This may not be acceptable in real applications, since deployed system equipped with certain defense mechanisms and only return final result (i.e., hard label) to client. In contrast, one possible approach formulating label attack, which can successfully executed within limited queries. To implement this idea, paper, we bypass reliance models benefit...
Abstract In generating adversarial examples, the conventional black-box attack methods rely on sufficient feedback from to-be-attacked models by repeatedly querying until is successful, which usually results in thousands of trials during an attack. This may be unacceptable real applications since Machine Learning as a Service Platform (MLaaS) only returns final result (i.e., hard-label) to client and system equipped with certain defense mechanisms could easily detect malicious queries. By...
Many attack techniques have been proposed to explore the vulnerability of DNNs and further help improve their robustness. Despite significant progress made recently, existing black-box methods still suffer from unsatisfactory performance due vast number queries needed optimize desired perturbations. Besides, other critical challenge is that adversarial examples built in a noise-adding manner are abnormal struggle successfully robust models, whose robustness enhanced by training against small...
In the rapidly evolving field of Artificial Intelligence Generated Content (AIGC), one key challenges is distinguishing AI-synthesized images from natural images. Despite remarkable capabilities advanced AI generative models in producing visually compelling images, significant discrepancies remain when these are compared to ones. To systematically investigate and quantify discrepancies, we introduce an AI-Natural Image Discrepancy Evaluation benchmark aimed at addressing critical question:...
Fondaparinux is a synthetic anticoagulant that inhibits thrombosis by suppressing factor Xa. The efficacy of fondaparinux for orthopedic surgeries has been revealed several foreign studies; however, relevant evidence in Chinese patients lacking. This study intended to investigate the occurrence rate and risk factors in-hospital venous thromboembolism (VTE), major bleeding, death receiving after surgery or trauma surgery.Totally, 1258 who received were retrospectively enrolled. Meanwhile,...
How to discriminate distal regulatory elements a gene target is challenging in understanding regulation and illustrating causes of complex diseases. Among known elements, enhancers interact with gene’s promoter regulate its expression. Although the emergence many machine learning approaches has been able predict enhancer-promoter interactions (EPIs), global precise prediction EPIs at genomic level still requires further exploration.In this paper, we develop an integrated method, called...