A5101647762- Topic Modeling
- Natural Language Processing Techniques
- Advanced Graph Neural Networks
- Speech and dialogue systems
- Data Quality and Management
- Advanced Neural Network Applications
- Climate variability and models
- Advanced MRI Techniques and Applications
- Image Retrieval and Classification Techniques
- MRI in cancer diagnosis
- Immune Response and Inflammation
- Systemic Lupus Erythematosus Research
- Mental Health Research Topics
- Atherosclerosis and Cardiovascular Diseases
- Service-Oriented Architecture and Web Services
- Multimodal Machine Learning Applications
- Cardiac Imaging and Diagnostics
- Text Readability and Simplification
- Birth, Development, and Health
- Time Series Analysis and Forecasting
- Adsorption and Cooling Systems
- Perfectionism, Procrastination, Anxiety Studies
- Brain Tumor Detection and Classification
- Cognitive Computing and Networks
- Plant Surface Properties and Treatments
Anhui Institute of Information Technology
2025
Inner Mongolia University of Technology
2024
XinHua Hospital
2024
Shanghai Jiao Tong University
2024
Xidian University
2021-2024
Nanchang University
2024
Jilin University
2011-2023
Chengdu University of Traditional Chinese Medicine
2022-2023
Zhuzhou Central Hospital
2022-2023
Xiangya Hospital Central South University
2023
The warming world greatly suffers the increase of frequency, severity and duration heat wave events, which would cause significant societal environmental damages implications from local to global scales. In order eliminate limitations site observations in accurate spatial extent identification large scale monitoring, this study tried investigate waves 2018 using datasets land surface temperature (LST) derived AQUA TIR sensors. A 15-year daily maximum LST dataset was used identify hot days...
Recent advances in Knowledge Graph Embedding (KGE) allow for representing entities and relations continuous vector spaces. Some traditional KGE models leveraging additional type information can improve the representation of which however totally rely on explicit types or neglect diverse representations specific to various relations. Besides, none existing methods is capable inferring all relation patterns symmetry, inversion composition as well complex properties 1-N, N-1 N-N relations,...
Natural language explanation in visual question answer (VQA-NLE) aims to explain the decision-making process of models by generating natural sentences increase users' trust black-box systems. Existing post-hoc methods have achieved significant progress obtaining a plausible explanation. However, such explanations are not always aligned with human logical inference, suffering from issues on: 1) Deductive unsatisfiability, generated do logically lead answer; 2) Factual inconsistency, model...
Open Information Extraction (OpenIE) facilitates domain-independent discovery of relational facts from large corpora. The technique well suits many open-world natural language understanding scenarios, such as automatic knowledge base construction, open-domain question answering, and explicit reasoning. Thanks to the rapid development in deep learning technologies, numerous neural OpenIE architectures have been proposed achieve considerable performance improvement. In this survey, we provide...
We introduce MiniMax-01 series, including MiniMax-Text-01 and MiniMax-VL-01, which are comparable to top-tier models while offering superior capabilities in processing longer contexts. The core lies lightning attention its efficient scaling. To maximize computational capacity, we integrate it with Mixture of Experts (MoE), creating a model 32 experts 456 billion total parameters, 45.9 activated for each token. develop an optimized parallel strategy highly computation-communication overlap...
Data augmentation is a cornerstone technique in deep learning, widely used to improve model generalization. Traditional methods like random cropping and color jittering, as well advanced techniques such CutOut, Mixup, CutMix, have achieved notable success across various domains. However, the mechanisms by which data improves generalization remain poorly understood, existing theoretical analyses typically focus on individual without unified explanation. In this work, we present framework that...
Speculative decoding accelerates inference in large language models (LLMs) by generating multiple draft tokens simultaneously. However, existing methods often struggle with token misalignment between the training and phases, limiting their performance. To address this, we propose GRIFFIN, a novel framework that incorporates token-alignable strategy model to mitigate misalignment. The employs loss masking mechanism exclude highly misaligned during training, preventing them from negatively...
Abstract Aiming at the complexity of glioma classification process, a framework based on SE‐ResNeXt network is proposed to simplify process benign and malignant gliomas. In addition, three optimization strategies are adopted improve accuracy. Firstly, MultiStepLR strategy used adjust learning rate dynamically in order ability network. Secondly, one‐hot label optimized by smoothing which can reduce dependence probability distribution real labels prediction Finally, transfer simplified CE‐MRI...
Open Information Extraction (OpenIE) facilitates domain-independent discovery of relational facts from large corpora. The technique well suits many open-world natural language understanding scenarios, such as automatic knowledge base construction, open-domain question answering, and explicit reasoning. Thanks to the rapid development in deep learning technologies, numerous neural OpenIE architectures have been proposed achieve considerable performance improvement. In this survey, we provide...
Systemic lupus erythematosus (SLE) is a systemic autoimmune disease involving multiple organs throughout the body. The health care-seeking behaviors, progression of SLE, and patients' knowledge attitudes toward SLE have not been characterized in China.The aim this study was to depict progression, medications patients with examine factors associated their flares, knowledge, China.We conducted cross-sectional survey 27 provinces China. Descriptive statistical methods were used demographic...
Additive manufacturing has significant advantages in complex parts of the vehicle manufacturing.As additive is a kind precise production activity, different components instruments need to be located appropriate positions ensure accuracy.The visual Simultaneous Localization and Mapping (SLAM) can considered practical means for this purpose.Considering dynamic characteristics scenarios, paper constructs deep learning-enhanced robust SLAM approach monitoring manufacturing.The proposed method...
The acute lung injury (ALI) mouse model induced by lipopolysaccharide (LPS) or endotoxin is still among the most commonly used models in animal studies of inflammation. current methods are an intraperitoneal injection LPS and tracheostomy for tracheal infusion LPS. However, former method lacks targeting damages other organs, latter induces operative trauma, infection risk, a low survival rate. Here, we recommend noninvasive oropharyngeal endotracheal intubation instillation mice. In this...
Background Since the end of 2019, Corona Virus Disease also known as COVID-19, has broken out in various countries. However, change China's COVID-19 prevention and control policy sharp increase number infected people are making teenagers have post-traumatic reactions. Negative reactions include: stress disorder (PTSD), depression, anxiety. Positive reaction mainly refers to growth (PTG). The purpose this study is explore reaction, which PTSD, anxiety co-occurrence pattern after trauma...
To investigate the feasibility and clinical utility of a compressed-sensing-accelerated subtractionless whole-body MRA (CS-WBMRA) protocol with only contrast injection for suspected arterial diseases, by comparison to conventional dual-pass subtraction-based (conventional-WBMRA) available computed tomography angiography (CTA).
Objective: To examine whether joint management of cancer pain by physicians and pharmacists in clinics provides economic advantages from the perspective Chinese healthcare system. Methods: From February 2018 to March 2020, 100 patients who visited clinic at Xiangya Hospital Central South University were included. These randomly assigned either control or intervention groups. The group received regular outpatient services a physician, while physician medication education provided pharmacist....
In multi-domain task-oriented dialogue systems, users proactively propose a series of domain-specific requests that can often be under-or over-specified, sometimes with ambiguous and cross-domain demands. System-sided initiative would necessary to identify certain situations appropriately interact resolve them. However, most existing systems fail consider such mixed-initiative interaction strategies, performing low efficiency poor collaboration ability in human-computer conversation. this...
Terrestrial ecosystem respiration (Reco) in drylands (arid and semi-arid areas) contributes to the largest uncertainty of global carbon cycle. Here, using Reco data from 24 sites (98 site-years) Fluxnet corresponding MODIS remote sensing products, we develop a novel semi-empirical, yet physiologically-based model: ILEP_Reco model (a derived ILEP, acronym for “integrated LE EVI proxy”). This can simulate observations across most biomes with small margin error (R 2 = 0.56, RMSE 1.12 gCm −2 d...
About 50 years ago, Chinese Great Famine (CGF) affected the entire population in China, and its long-term influence on offspring has attracted significant attention for research. However, information possible metabolic differences between sexes is limited. This study explored whether there might be sex risks of development glucolipid dysfunction fatty liver following prenatal exposure to CGF.There were 11,417 subjects around 55 age (6,661 women 4,756 men). They divided as exposed group which...
Building document-grounded dialogue systems have received growing interest as documents convey a wealth of human knowledge and commonly exist in enterprises. Wherein, how to comprehend retrieve information from is challenging research problem. Previous work ignores the visual property treats them plain text, resulting incomplete modality. In this paper, we propose Layout-aware document-level Information Extraction dataset, LIE, facilitate study extracting both structural semantic visually...