A5046612317- Face and Expression Recognition
- Anomaly Detection Techniques and Applications
- Domain Adaptation and Few-Shot Learning
- Underwater Acoustics Research
- Sparse and Compressive Sensing Techniques
- Geophysical Methods and Applications
- Machine Learning and ELM
- Advanced SAR Imaging Techniques
- Advanced Computing and Algorithms
- Physics of Superconductivity and Magnetism
- Speech Recognition and Synthesis
- Graphene research and applications
- Energy Load and Power Forecasting
- Image and Video Stabilization
- Computer Graphics and Visualization Techniques
- Electric Power System Optimization
- Machine Learning in Healthcare
- Music and Audio Processing
- Perovskite Materials and Applications
- Human Pose and Action Recognition
- Smart Grid and Power Systems
- Spider Taxonomy and Behavior Studies
- Speech and Audio Processing
- Wireless Signal Modulation Classification
- Simulation and Modeling Applications
Inner Mongolia University of Science and Technology
2020-2022
University of Electronic Science and Technology of China
2005-2021
Tencent (China)
2021
Nanjing University of Posts and Telecommunications
2021
National Engineering Research Center of Electromagnetic Radiation Control Materials
2020
Mongolian University of Science and Technology
2013
Auto-Encoder (AE)-based deep subspace clustering (DSC) methods have achieved impressive performance due to the powerful representation extracted using neural networks while prioritizing categorical separability. However, self-reconstruction loss of an AE ignores rich useful relation information and might lead indiscriminative representation, which inevitably degrades performance. It is also challenging learn high-level similarity without feeding semantic labels. Another unsolved problem...
Multiple kernel learning (MKL) method is generally believed to perform better than single method. However, some empirical studies show that this not always true: the combination of multiple kernels may even yield an worse performance using a kernel. There are two possible reasons for failure: (i) most existing MKL methods assume optimal linear base kernels, which hold true; and (ii) weights inappropriately assigned due noises carelessly designed algorithms. In paper, we propose novel...
Valley pseudospin in two-dimensional (2D) transition-metal dichalcogenides (TMDs) allows optical control of spin-valley polarization and intervalley quantum coherence. Defect states TMDs give rise to new exciton features theoretically exhibit polarization; however, experimental achievement this phenomenon remains challenges. Here, we report unambiguous valley defect-bound localized excitons CVD-grown monolayer MoS2; enhanced Zeeman splitting with an effective g-factor -6.2 is observed. Our...
Synthetic aperture radar automatic target recognition (SAR ATR) is a key technique of remote-sensing image recognition, which has many potential applications in the fields military surveillance, national defense, civil application, and so on. With development science technology, deep convolutional neural network (DCNN) been widely applied for SAR ATR. However, it difficult to use learning train models with limited ray images. To resolve this problem, we proposed an effectively lightweight...
Multiple kernel learning (MKL) method is generally believed to perform better than single method. However, some empirical studies show that this not always true: the combination of multiple kernels may even yield an worse performance using a kernel. There are two possible reasons for failure: (i) most existing MKL methods assume optimal linear base kernels, which hold true; and (ii) weights inappropriately assigned due noises carelessly designed algorithms. In paper, we propose novel...
High T <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">c</sub> superconducting (HTS) cables and their application to build a DC power transmission system with the advantages of high transport current capability no resistive loss have been studied. Technical assessments electrical behaviors use HTS cable technology carried out. The its models built, analysis results obtained mainly include magnetic field distribution performances.
The problem of unknown modulation signal recognition has been received intensely attentions in next-generational intelligent wireless communications. deep learning (DL) widely used due to its excellent performance solving classification problems and the DL-based automatic (AMC) had proposed. However, AMC method usually high space complexity computational complexity, which limits miniaturized devices with limited storage computing capability. Therefore, a lightweight residual neural network...
Singer recognition aims to automatically recognize the singer of a given recording. Compared spoken voices, singing voice is characterized by much higher degree vocal style. The task becomes more challenging when it operates on numerous singers. This paper explores different strategies in deep metric learning framework, with special focus their performance large-scale dataset consisting audio samples from 5057 We conduct thorough experiments compare loss functions, including triplet loss,...
Virtual Endoscopy is a new non-invasive inspection method, it gets the body's two-dimensional slice data by CT, MRI and collects volume to generate three-dimensional model of various organs body, then generates visual display, glimpses roaming function. This article imitates camera visualization toolkit, using flexible keyboard mouse control virtual roam in trachea, having achieved endoscopy. endoscopy has broad application prospects realization medical diagnostic therapy teaching.
Leveraging on the underlying low-dimensional structure of data, low-rank and sparse modeling approaches have achieved great success in a wide range applications. However, many applications data can display structures beyond simply being or sparse. Fully extracting exploiting hidden information is always desirable favorable. To reveal more effective manifold structure, this paper, we explicitly model relation. Specifically, propose learning framework that retains pairwise similarities between...
Deep auto-encoders (DAEs) have achieved great success in learning data representations via the powerful representability of neural networks. But most DAEs only focus on dominant structures which are able to reconstruct from a latent space and neglect rich structural information. In this work, we propose new representation method that explicitly models leverages sample relations, turn is used as supervision guide learning. Different previous our framework well preserves relations between...