A5034556928- Natural Language Processing Techniques
- Speech Recognition and Synthesis
- Topic Modeling
- Gas Sensing Nanomaterials and Sensors
- Music and Audio Processing
- MRI in cancer diagnosis
- 3D Shape Modeling and Analysis
- Computer Graphics and Visualization Techniques
- Text Readability and Simplification
- Domain Adaptation and Few-Shot Learning
- Analytical Chemistry and Sensors
- Multimodal Machine Learning Applications
- Advanced Vision and Imaging
- Advanced MRI Techniques and Applications
- Iron Metabolism and Disorders
- Prenatal Screening and Diagnostics
- Human Pose and Action Recognition
- Hemoglobinopathies and Related Disorders
- Advanced Chemical Sensor Technologies
- Genomic variations and chromosomal abnormalities
- Advanced Neural Network Applications
- Speech and dialogue systems
- Advanced Neuroimaging Techniques and Applications
- Infrared Thermography in Medicine
- Glioma Diagnosis and Treatment
Huazhong University of Science and Technology
2021-2025
Tongji Hospital
2021-2025
Microsoft (United States)
2023-2024
Gansu Provincial Maternal and Child Health Hospital
2018-2024
University of Illinois Urbana-Champaign
2022-2024
Fujian Medical University
2023-2024
Central South University of Forestry and Technology
2022-2024
Central South University
2022-2024
Fondazione Bruno Kessler
2024
University of Trento
2024
We present a new large-scale multilingual video description dataset, VATEX <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup> , which contains over 41,250 videos and 825, 000 captions in both English Chinese. Among the captions, there are 206,000 English-Chinese parallel translation pairs. Compared to widely-used MSRVTT dataset [64], is multilingual, larger, linguistically complex, more diverse terms of natural language descriptions. also...
The CoPM-24 gas sensor material showed an excellent sensing performance to nitrogen oxides, that mainly attributed their adsorption property and unique heterostructure.
Semantic composition functions have been playing a pivotal role in neural representation learning of text sequences. In spite their success, most existing models suffer from the underfitting problem: they use same shared compositional function on all positions sequence, thereby lacking expressive power due to incapacity capture richness compositionality. Besides, different tasks are independent and learned scratch. this paper, we propose new sharing scheme across multiple tasks....
This paper studies learning logic rules for reasoning on knowledge graphs. Logic provide interpretable explanations when used prediction as well being able to generalize other tasks, and hence are critical learn. Existing methods either suffer from the problem of searching in a large search space (e.g., neural programming) or ineffective optimization due sparse rewards techniques based reinforcement learning). To address these limitations, this proposes probabilistic model called RNNLogic....
This paper proposes NeuralEditor that enables neural radiance fields (NeRFs) natively editable for general shape editing tasks. Despite their impressive results on novel-view synthesis, it remains a fundamental challenge NeRFs to edit the of scene. Our key insight is exploit explicit point cloud representation as underlying structure construct NeRFs, inspired by intuitive interpretation NeRF rendering process projects or "plots" associated 3D 2D image plane. To this end, introduces novel...
Whitfordiodendron filipes var. tomentosum is an endemic plant in China. There have been no chemical or pharmacological studies of this reported before. In the current research, eight triterpenes and two steroids were obtained. Their structures established by analysis NMR data comparison with those literature. These ten structurally diverse compounds comprised five distinct carbon frameworks different functionalities. The chemotaxonomic significance these secondary metabolites was discussed,...
In this work, the obtained C-ZIF-67/PAN <italic>via</italic> one-step pyrolysis of hybrid ZIF-67/PAN nanofibers were used to fabricate electrochemical sensor for determination benzenediol isomer.
Machine learning has huge potential to revolutionize the field of drug discovery and is attracting increasing attention in recent years. However, lacking domain knowledge (e.g., which tasks work on), standard benchmarks data preprocessing pipelines are main obstacles for machine researchers this domain. To facilitate progress discovery, we develop TorchDrug, a powerful flexible platform built on top PyTorch. TorchDrug variety important including molecular property prediction, pretrained...
Variants dropout methods have been designed for the fully-connected layer, convolutional layer and recurrent in neural networks, shown to be effective avoid overfitting. As an appealing alternative layers, self-attention surprisingly lacks a specific method. This paper explores possibility of regularizing attention weights Transformers prevent different contextualized feature vectors from co-adaption. Experiments on wide range tasks show that DropAttention can improve performance reduce
Neonatal sepsis, a severe infectious disease associated with high mortality rates, is characterized by metabolic disturbances that play crucial role in its progression. The aim of this study to develop metabolism-related model for assessing 30-day neonatal sepsis. clinical data sepsis at Ganzhou Women and Children's Health Care Hospital from January 2019 December 2022 were retrospectively analyzed. cases divided into survival non-survival groups. Multivariate logistic regression analysis was...
Streaming multi-talker speech translation is a task that involves not only generating accurate and fluent translations with low latency but also recognizing when speaker change occurs what the speaker's gender is. Speaker information can be used to create audio prompts for zero-shot text-to-speech system, help select profiles in conventional model. We propose tackle streaming detection classification by incorporating embeddings into transducer-based end-to-end Our experiments demonstrate...
Semantic composition functions have been playing a pivotal role in neural representation learning of text sequences. In spite their success, most existing models suffer from the underfitting problem: they use same shared compositional function on all positions sequence, thereby lacking expressive power due to incapacity capture richness compositionality. Besides, different tasks are independent and learned scratch. this paper, we propose new sharing scheme across multiple tasks....
Synthesis of hierarchical mixed phase WO<sub>3</sub> by effectively utilizing the structure hemp. The morphology and synergistic effect improved NO<sub>2</sub> gas sensitivity performance at RT.
Simultaneous speech-to-text translation is widely useful in many scenarios.The conventional cascaded approach uses a pipeline of streaming ASR followed by simultaneous MT, but suffers from error propagation and extra latency.To alleviate these issues, recent efforts attempt to directly translate the source speech into target text simultaneously, this much harder due combination two separate tasks.We instead propose new paradigm with advantages both endto-end approaches.The key idea use...
Chromosomal abnormalities are the main cause of birth defects in newborns. Since inception noninvasive prenatal testing (NIPT) technology, it has primarily been applied to detection common trisomy (T21, T18, T13). However, application NIPT microdeletion and microduplication is still controversial.