A5100346983- Domain Adaptation and Few-Shot Learning
- Advanced Neural Network Applications
- Multimodal Machine Learning Applications
- Anomaly Detection Techniques and Applications
- Natural Language Processing Techniques
- Adversarial Robustness in Machine Learning
- Topic Modeling
- Advanced Malware Detection Techniques
- Machine Learning and Data Classification
- Advanced Text Analysis Techniques
- Mineral Processing and Grinding
- Robotic Path Planning Algorithms
- Salmonella and Campylobacter epidemiology
- Internet Traffic Analysis and Secure E-voting
- Physical Unclonable Functions (PUFs) and Hardware Security
- Brain Tumor Detection and Classification
- Digital Imaging for Blood Diseases
- Geochemistry and Geologic Mapping
- Robot Manipulation and Learning
- Complexity and Algorithms in Graphs
- Remote-Sensing Image Classification
- Cooperative Communication and Network Coding
- Medical Image Segmentation Techniques
- Hydrocarbon exploration and reservoir analysis
- EEG and Brain-Computer Interfaces
Shanghai Pudong New Area Gongli Hospital
2023-2024
Second Military Medical University
2023-2024
Institute of Automation
2024
Chinese Academy of Sciences
2024
Texas A&M University
2010-2023
Xiamen University
2023
Wuchang Shouyi University
2022
Intel (United Kingdom)
2022
University of Jinan
2022
Wuchang University of Technology
2022
Campylobacter jejuni (C. jejuni) is one of the most common causes human bacterial enteritis worldwide primarily due to contaminated poultry products. Previously, we found a significant difference in C. colonization ceca between two genetically distinct broiler lines (Line A (resistant) has less colony than line B (susceptible) on day 7 post inoculation). We hypothesize that different mechanisms these genetic may affect their ability resist chickens. The molecular local host response chickens...
Semi-supervised dense prediction tasks, such as semantic segmentation, can be greatly improved through the use of contrastive learning. However, this approach presents two key challenges: selecting informative negative samples from a highly redundant pool and implementing effective data augmentation. To address these challenges, we present an adversarial learning method specifically for semi-supervised segmentation. Direct negatives is adopted to retain discriminative information past,...
Recently, adversarial attacks have shown to lead the state-of-the-art deep neural networks (DNNs) misclassification. However, most are generated according whether they perceptual human visual system, measured by geometric metrics such as <inline-formula><tex-math notation="LaTeX">$\ell _2$</tex-math></inline-formula> -norm, which ignores common watermarks in cyber-physical systems. In this article, we propose a fast watermark attack (FAWA) method based on differential evolution technique,...
Vision-language models have recently shown great potential on many tasks in computer vision. Meanwhile, prior work demonstrates prompt tuning designed for vision-language could acquire superior performance few-shot image recognition compared to linear probe, a strong baseline. In practice, are inherently correlated, particularly within specialized domains. However, such information is overlooked previously. Inspired by the fact that modeling task relationship multi-task learning can usually...
This work studies the Stopping-Set Elimination Problem, namely, given a stopping set, how to remove fewest erasures so that remaining can be decoded by belief propagation in k iterations (including = ∞). The NP-hardness of problem is proven. An approximation algorithm presented for 1. And efficient exact algorithms are general when sets form trees.
Training a generalized reliable model is great challenge since sufficiently labeled data are unavailable in some open application scenarios. Few-shot learning (FSL) aims to learn new problems with only few examples that can tackle this problem and attract extensive attention. This paper proposes novel few-shot method based on double pooling squeeze excitation attention (dSE) for the purpose of improving discriminative ability by proposing feature expression. Specifically, proposed dSE module...
This paper uses an encoder-decoder framework based on semantic to feature analysis construct a neural machine translation model, let the automatically perform learning, transform corpus data into word vectors in distributed representation, and use networks implement source language Direct mapping between target languages. A alignment method deep network is proposed, which effectively utilizes similarity of vocabulary context information model more accurately. dimensionality reduction learn...
Sophisticated traffic analytics, such as the encrypted analytics and unknown malware detection, emphasizes need for advanced methods to analyze network traffic.Traditional of using fixed patterns, signature matching, rules detect known patterns in are being replaced with AI (Artificial Intelligence) driven algorithms.However, absence a high-performance networking-specific framework makes deploying real-time AI-based processing within networking workloads impossible.In this paper, we describe...
Unsupervised learning of vision transformers seeks to pretrain an encoder via pretext tasks without labels. Among them is the Masked Image Modeling (MIM) aligned with pretraining language by predicting masked patches as a task. A criterion in unsupervised task needs be sufficiently hard prevent transformer from trivial low-level features not generalizable well downstream tasks. For this purpose, we propose Adversarial Positional Embedding (AdPE) approach -- It distorts local visual...
With the increasing availability of datasets exhibiting fine granularity and subtle differences between categories, fine-grained visual categorization tasks have gained significant attention across various domains. However, focus often lies solely on overall dataset performance metrics such as top-l accuracy, while lacking a comprehensive understanding underlying factors. This paper addresses this gap by presenting detailed analysis CUB-200-2011 through extensive experiments. We identify...
Background: The oil-water relative permeability curve is an important experimental data and basis for oilfield development scheme dynamic prediction. characteristics of oil water curves in different reservoirs are otherness. Objective: In order to enable various effects be reflected the standardized curve, can reflect each core. Method: core taken from wells a similar flow unit. measured by indoor physical simulation experiments. representativeness conforming hydrodynamic screened. Because...
DNN-based video object detection (VOD) powers autonomous driving and surveillance industries with rising importance promising opportunities. However, adversarial patch attack yields huge concern in live vision tasks because of its practicality, feasibility, powerful effectiveness. This work proposes Themis, a software/hardware system to defend against patches for real-time robust detection. We observe that exhibit extremely localized superficial feature small region non-robust predictions,...
Label noise and class imbalance commonly coexist in real-world data. Previous works for robust learning, however, usually address either one type of the data biases underperform when facing them both. To mitigate this gap, work presents a novel meta-learning based dynamic loss that automatically adjusts objective functions with training process to robustly learn classifier from long-tailed noisy Concretely, our comprises label corrector margin generator, which respectively correct labels...