A5101578575- Environmental Changes in China
- Computational Drug Discovery Methods
- Protein Structure and Dynamics
- Speech and Audio Processing
- Speech Recognition and Synthesis
- Advancements in Battery Materials
- Advanced Battery Materials and Technologies
- Bayesian Modeling and Causal Inference
- Machine Learning in Materials Science
- Sparse and Compressive Sensing Techniques
- Natural Language Processing Techniques
- Statistical Methods and Bayesian Inference
- Advanced battery technologies research
- Face recognition and analysis
- Heat Transfer Mechanisms
- Hydrological Forecasting Using AI
- Intracranial Aneurysms: Treatment and Complications
- Coastal and Marine Management
- Technology and Security Systems
- Remote Sensing and Land Use
- Advanced Computational Techniques and Applications
- Fluid Dynamics and Turbulent Flows
- Multi-Criteria Decision Making
- Advanced Image Processing Techniques
- Surface Roughness and Optical Measurements
Xi'an University of Technology
2024-2025
Handan College
2023-2024
Xinjiang Agricultural University
2024
Shanghai Jiao Tong University
2019-2023
Yunnan University
2020
University of Toronto
2017
National Marine Environmental Monitoring Center
2008
China National Environmental Monitoring Center
2008
South China University of Technology
2007
Protein binding site prediction is an important prerequisite task of drug discovery and design. While sites are very small, irregular varied in shape, making the challenging. Standard 3D U-Net has been adopted to predict but got stuck with unsatisfactory results, incomplete, out-of-bounds, or even failed. The reason that this scheme less capable extracting chemical interactions entire region hardly takes into account difficulty segmenting complex shapes. In paper, we propose a refined...
It is critical to accurately predict the rupture risk of an intracranial aneurysm (IA) for timely and appropriate treatment because fatality rate after 50%. Existing methods relying on morphological features (e.g., height-width ratio) measured manually by neuroradiologists are labor intensive have limited use assessment. Therefore, we propose end-to-end deep-learning method, called TransIAR net, automatically learn from 3D computed tomography angiography (CTA) data status IA rupture. We...
Protein binding site prediction is an important prerequisite for the discovery of new drugs. Usually, natural 3D U-Net adopted as standard framework to do per-voxel binary mask classification. However, this scheme only performs feature extraction single-scale samples, which may bring loss global or local information, resulting in incomplete, artifacted even missed predictions. To tackle issue, we propose a network called GLPocket, based on Lmser (Least mean square error reconstruction) and...
Abstract Flood forecast models have become better through research as they led to a lower risk of flooding, policy ideas, less human death, and destruction property, so this study uses Scientometric analysis for floods. In analysis, citation-based data are used uncover major publishing areas, such the most prominent keywords, top best commonly publications, highly cited journal articles, countries, authors that achieved consequent distinction in flood analysis. Machine learning (ML)...
It is an essential step to locate the binding sites or pockets of drug molecules on protein structure in design. This challenging because 3D structures are usually complicated, irregular shape and relatively small. Existing deep learning methods for this task U-Net models, they have forward skip connections efficiently transfer features different levels from encoder decoder improving pocket prediction. However, there still room improve prediction accuracy. In paper, we propose RecurPocket, a...
Methods are developed for eliciting a Dirichlet prior based upon stating bounds on the individual probabilities that hold with high probability. This approach to selecting is applied contingency table problem where it demonstrated how assess respect bias induces as well check prior-data conflict. It shown assessment of hypothesis via relative belief can easily take into account what means falsity correspond difference practical importance and provide evidence in favor hypothesis.
The "shuttle effect" and several issues related to it are seen as "obstacles" the study development of lithium-sulfur batteries (LSBs). This work aims at finding how fully expose bimetallic sites quicken battery reaction kinetics. Here, a NiCo-MOF its derivative NiCo@C with hollow sea urchin structure produced. obtained possesses micromesoporous disclosed active because distinctive structure. experimental findings demonstrate that exposed take on chemical adsorbents collaborate...
Existing single image super-resolution (SISR) methods usually focus on Low-Resolution (LR) images which are artificially generated from High-Resolution (HR) by a down-sampling process, but not robust for unmatched training set and testing set. This paper proposes GAN Flexible Lmser (GFLmser) network that bidirectionally learns the High-to-Low (H2L) process degrades HR to LR Low-to-High (L2H) recovers back images. The two directions share same architecture, added with gated skip connections...
The Seriphidium transiliense desert pasture is an important spring-autumn in northern Xinjiang, China, and has been subjected to grazing by livestock at different intensities, thus resulting widespread deterioration of its biodiversity ecosystem services. To understand the response mechanism stoichiometric characteristics vegetation grazing, leaf carbon (C), nitrogen (N), phosphorus (P) C : N P ratios S. were studied under intensities. results show that control C, contents N, 458.79 ± 53.5...
In this paper, we propose a Myanmar speech synthesis system based on an End-to-End neural network model, which integrates the phone model into Tacotron2 model. Based Seq2seq architecture, use phone-level embedding to form feature prediction from sequences Mel spectrum, and combine with semi-supervised generation generate high-quality synthesized speech. addition, introduced BERT pre-training decoder module assist extraction, reduces system's dependence extraction improve text richness....
Current face recognition tasks are usually carried out on high-quality images, but in reality, most images captured under unconstrained or poor conditions, e.g., by video surveillance. Existing methods featured learning data uncertainty to avoid overfitting the noise, adding margins angle cosine space of normalized softmax loss penalize target logit, which enforces intra-class compactness and inter-class discrepancy. In this paper, we propose a deep Rival Penalized Competitive Learning...
Methods are developed for eliciting a Dirichlet prior based upon bounds on the individual probabilities that hold with virtual certainty. This approach to selecting is applied contingency table problem where it demonstrated how assess bias in as well check prior-data conflict. It shown assessment of hypothesis via relative belief can easily take into account what means falsity correspond difference practical importance and provide evidence favor hypothesis.
Myanmar belongs to the Lolo-Burmese sub-branch of Tibeto-Burmese branch Sino-Tibetan language family and is a tonal language. In front-end text analysis speech synthesis, prosodic structure unit's boundary prediction are crucial naturalness synthesis. order improve this paper studies features unit prediction. The size units duration syllables before after their boundaries have been studied in paper. To realize automatic labeling, method labeling based on combination word segmentation silence...