A5100413330- Topic Modeling
- Natural Language Processing Techniques
- Multimodal Machine Learning Applications
- Speech and dialogue systems
- Video Surveillance and Tracking Methods
- Image and Video Quality Assessment
- Domain Adaptation and Few-Shot Learning
- Advanced Neural Network Applications
- Parallel Computing and Optimization Techniques
- Text and Document Classification Technologies
- Visual Attention and Saliency Detection
- Human Pose and Action Recognition
- Anomaly Detection Techniques and Applications
- Advanced Graph Neural Networks
- Image Enhancement Techniques
- Speech Recognition and Synthesis
- Advanced Text Analysis Techniques
- Advanced Image Processing Techniques
- Neural Networks and Applications
- Rough Sets and Fuzzy Logic
- Embedded Systems Design Techniques
- Advanced Image and Video Retrieval Techniques
- Advanced Decision-Making Techniques
- Advanced Computational Techniques and Applications
- Time Series Analysis and Forecasting
Southwest Hospital
2024-2025
Xiangya Hospital Central South University
2025
Union Hospital
2022-2025
Huazhong University of Science and Technology
2022-2025
China Meteorological Administration
2025
Southwest General Health Center
2024
National Institute of Metrology
2024
Inner Mongolia University
2008-2024
China National Administration of Coal Geology
2024
Shenyang University
2024
Tuning pre-trained language models (PLMs) with task-specific prompts has been a promising approach for text classification. Particularly, previous studies suggest that prompt-tuning remarkable superiority in the low-data scenario over generic fine-tuning methods extra classifiers. The core idea of is to insert pieces, i.e., template, input and transform classification problem into masked modeling problem, where crucial step construct projection, verbalizer, between label space word space. A...
Recommender system usually suffers from severe <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">popularity bias</i> — the collected interaction data exhibits quite imbalanced or even long-tailed distribution over items. Such skewed may result users' xmlns:xlink="http://www.w3.org/1999/xlink">conformity</i> to group, which deviates reflecting true preference. Existing efforts for tackling this issue mainly focus on completely eliminating...
User-generated content (UGC) live videos are often bothered by various distortions during capture procedures and thus exhibit diverse visual qualities. Such source further compressed transcoded media server providers before being distributed to end-users. Because of the flourishing UGC videos, effective video quality assessment (VQA) tools needed monitor perceptually optimize streaming in distributing process. In this paper, we address Live VQA problems constructing a first-of-a-kind...
We present a GAN-based Transformer for general action-conditioned 3D human motion generation, including not only single-person actions but also multi-person interactive actions. Our approach consists of powerful Action-conditioned TransFormer (ActFormer) under GAN training scheme, equipped with Gaussian Process latent prior. Such design combines the strong spatio-temporal representation capacity Transformer, superiority in generative modeling GAN, and inherent temporal correlations from...
The Kalman filter has been adopted in acoustic echo cancellation due to its robustness double-talk, fast convergence, and good steady-state performance. performance of is closely related the estimation accuracy state noise covariance observation covariance. error may lead unacceptable results, especially when path suffers abrupt changes, tracking could be degraded significantly. In this paper, we propose neural filtering (NKF), which uses networks implicitly model output gain real-time....
This paper presents a new robotic navigation aid, called Co-Robotic Cane (CRC). The CRC uses 3D camera for both pose estimation and object recognition in an unknown indoor environment. 6-DOF method determines the CRC's change by egomotion iterative closest point algorithm reduces integration error graph optimization algorithm. does not require any prior knowledge of detects structures such as stairways, doorways, etc. objects tables, computer monitors, Gaussian Mixture Model based pattern...
Recently, prompt-based methods have achieved significant performance in few-shot learning scenarios by bridging the gap between language model pre-training and fine-tuning for downstream tasks. However, existing prompt templates are mostly designed sentence-level tasks inappropriate sequence labeling objectives. To address above issue, we propose a multi-task instruction-based generative framework, named InstructionNER, low-resource entity recognition. Specifically, reformulate NER task as...
Abstract We explored a new artificial intelligence-assisted method to assist junior ultrasonographers in improving the diagnostic performance of uterine fibroids and further compared it with senior confirm effectiveness feasibility intelligence method. In this retrospective study, we collected total 3870 ultrasound images from 667 patients mean age 42.45 years ± 6.23 [SD] for those who received pathologically confirmed diagnosis 570 women 39.24 5.32 without lesions Shunde Hospital Southern...
Osteosarcoma, a malignant bone tumor that occurs in adolescents, proliferates and is prone to pulmonary metastasis. Osteosarcoma characterized by high genotypic heterogeneity, making it difficult identify reliable anti‐osteosarcoma targets. The genotype of osteosarcoma may be highly dynamic, but its dependence on energy remains constant. Fortunately, tumors tend have relatively consistent metabolic types. Targeting metabolism with anti‐tumor therapies new strategy for treating tumors. Genes...
Abstract In recent years, the Weddell Sea ice loss accounted for nearly half of Antarctic sea loss, whose prediction has attracted a growing number attentions. this study, we developed two types monthly spatial statistical models concentration (SIC) in Sea, specifically Multiple Linear Regression (MLR) and Multivariate Empirical Orthogonal Function (MEOF) models. Both shared common set advanced oceanic atmospheric variables, including Pacific Decadal Oscillation, eastern tropical ocean...
Abstract Background Accurate detection and grading of fresh rib fractures are crucial for patient management but remain challenging due to the complexity structures on CT images. Methods Chest images from 383 patients with were retrospectively analyzed. The dataset was divided into a training set ( n = 306) an internal testing 77). An external 50 public RibFrac included. Fractures classified severe non-severe categories. A modified YOLO-based deep learning model developed grading....
The frequency-hopping communication system has been widely used in anti-jamming due to its anti-interception and performance. With the increasingly complex electromagnetic environment, needs more flexible patterns deal with interferences, which brings great challenges receiver. In this paper, an intelligent receiving scheme of sequences is proposed, combines time–frequency analysis deep learning realize estimation sequences. A hybrid network module designed by combining a convolutional...
Prompt tuning learns soft prompts to condition the frozen Pre-trained Language Models (PLMs) for performing downstream tasks in a parameter-efficient manner. While prompt has gradually reached performance level of fine-tuning as model scale increases, there is still large gap between and models moderate small scales (typically less than 11B parameters). In this paper, we empirically show that trained tokens can have negative impact on task thus degrade its performance. To bridge gap, propose...
Meta-learning methods have been widely used in few-shot named entity recognition (NER), especially prototype-based methods. However, the Other(O) class is difficult to be represented by a prototype vector because there are generally large number of samples that miscellaneous semantics. To solve problem, we propose MeTNet, which generates vectors for types only but not O-class. We design an improved triplet network map and into low-dimensional space easier classified adaptive margin each...
In this letter, a method called Semantic Information Oriented No-Reference (SIONR) video quality assessment model is developed, which can effectively represent degradation of by taking the variations semantic information into consideration. Specially, temporal features between adjacent frames are calculated to consider inconsistency static information. Moreover, low-level also applied as supplementary take distortions related local details Experimental results demonstrate that our proposed...
Yutao Mou, Keqing He, Yanan Wu, Zhiyuan Zeng, Hong Xu, Huixing Jiang, Wei Weiran Xu. Proceedings of the 60th Annual Meeting Association for Computational Linguistics (Volume 2: Short Papers). 2022.
Fusing hand-based features in multi-modal biometric recognition enhances anti-spoofing capabilities.Additionally, it leverages inter-modal correlation to enhance performance.Concurrently, the robustness and performance of system can be enhanced through judiciously leveraging among multimodal features.Nevertheless, two issues persist feature fusion recognition: Firstly, enhancement has not comprehensively considered inter-modality correlations distinct modalities.Secondly, during modal...
Xin Mao, Meirong Ma, Hao Yuan, Jianchao Zhu, ZongYu Wang, Rui Xie, Wei Wu, Man Lan. Proceedings of the 60th Annual Meeting Association for Computational Linguistics (Volume 1: Long Papers). 2022.