A5100381069- Advanced Neural Network Applications
- Electromagnetic Scattering and Analysis
- Robotic Path Planning Algorithms
- Advanced Memory and Neural Computing
- Advanced Antenna and Metasurface Technologies
- Electromagnetic Simulation and Numerical Methods
- Multimodal Machine Learning Applications
- GNSS positioning and interference
- Soft Robotics and Applications
- Human Pose and Action Recognition
- Solidification and crystal growth phenomena
- Indoor and Outdoor Localization Technologies
- Anomaly Detection Techniques and Applications
- Inertial Sensor and Navigation
- Metamaterials and Metasurfaces Applications
- Control and Dynamics of Mobile Robots
- Domain Adaptation and Few-Shot Learning
- Advanced Numerical Methods in Computational Mathematics
- CCD and CMOS Imaging Sensors
- Differential Equations and Numerical Methods
- Advanced Mathematical Modeling in Engineering
- Adversarial Robustness in Machine Learning
- Numerical methods for differential equations
- Advanced Image and Video Retrieval Techniques
- Antenna Design and Analysis
Northeastern University
2022-2025
Institute of High Performance Computing
2021-2025
Shenzhen University
2024
Second Hospital of Shandong University
2024
Chinese Academy of Medical Sciences & Peking Union Medical College
2024
Southern University of Science and Technology
2023-2024
Chinese University of Hong Kong, Shenzhen
2024
Beijing Normal University - Hong Kong Baptist University United International College
2024
Sun Yat-sen University
2021-2024
The Seventh Affiliated Hospital of Sun Yat-sen University
2024
The image of a face varies with the illumination, pose, and facial expression, thus we say that single is high uncertainty for representing face. In this sense, just an observation it should not be considered as absolutely accurate representation As more images from same person provide observations face, may useful reducing improving accuracy recognition. However, in real world recognition system, subject usually has only limited number available there uncertainty. paper, attempt to improve...
Temporal action localization (TAL), which involves recognizing and locating instances, is a challenging task in video understanding. Most existing approaches directly predict classes regress offsets to boundaries, while overlooking the discrepant importance of each frame. In this paper, we propose an Action Sensitivity Learning framework (ASL) tackle task, aims assess value frame then leverage generated sensitivity recalibrate training procedure. We first introduce lightweight Evaluator...
The unprecedented development of Internet Things results in the explosion spatiotemporal signals generated by smart edge devices, leading to a surge interest real‐time learning such data. This imposes big challenge conventional digital hardware because physically separated memory and processing units transistor scaling limit. Memristors are deemed solution for efficient portable deep learning. However, their ionic resistive switching incurs large programming stochasticity energy,...
The availability of GNSS carrier-phase measurements from smartphones and tablets has made accurate positioning possible with smartphones. However, it requires the development advanced algorithms to process In this research, we propose use time-differenced (TDCP) observations, instead raw Doppler observations improve precise point (PPP) real-time kinematic (RTK) performance Although can contribute velocity estimation a moving object subsequently solutions, current Android are found biased...
Artificial intelligence (AI) techniques have been spreading in most scientific areas and become a heated focus photonics research recent years. Forward modeling inverse design using AI can achieve high efficiency accuracy for components. With AI-assisted electronic circuit components, more advanced applications emerged. Photonics benefit great deal from AI, turn, benefits by carrying out algorithms, such as complicated deep neural networks components that use photons rather than electrons....
Cross-media retrieval of scientific and technological information is one the important tasks in cross-media study. obtain target from massive multi-source heterogeneous resources, which helps to design applications that meet users' needs, including recommendation, personalized retrieval, etc. The core learn a common subspace, so data different media can be directly compared with each other after being mapped into this subspace. In subspace learning, existing methods often focus on modeling...
In this work, a full-wave partial element equivalent circuit (PEEC) model for 2-D periodic metallic structures is proposed. The mixed potential integral equations (MPIEs) with Green’s function (GF) are interpreted as loop following Kirchhoff’s voltage law (KVL) in domain. impact of gauge freedom the MPIE on corresponding PEEC discussed. A tailored introduced to form generalized structures, which demonstrates superior passivity and frequency stability. With proposed model, lumped active...
Metabolic pathway analysis, one of the most important fields in biochemistry, is pivotal to understanding maintenance and modulation functions an organism.Good comprehension metabolic pathways critical mechanisms some fundamental biological processes.Given a small molecule or enzyme, how may identify which it participate?Answering such question first step system.By utilizing information provided by chemical-chemical interactions, chemical-protein protein-protein novel method was proposed...
This paper presents a quasi-static surface-based partial element equivalent circuit (PEEC) model for electromagnetic problems consisting of coupled conductors and heterogeneous dielectrics finite size. The PEEC is based on the surface principle. Unlike traditional models, in which integral equations are set up by enforcing field continuity true region, homogeneous this obtained tangential null condition regions. Simplified under assumption, elements carrying electric currents magnetic...
In this article, a passive full-wave partial element equivalent circuit (PEEC) model is formulated for modeling electromagnetic (EM) problems with finite dielectrics. The proposed PEEC substantial extension of the generalized multiple conductors in free space. reflects physical radiation mechanism an EM problem effects, both from and dielectrics, are incorporated real-valued frequency-dependent inductances resistances, while capacitances remain to be frequency invariant. Consequently,...
Abstract Pedestrian detection is mostly used in autonomous driving scenarios, which require high real-time detection. Most of the existing algorithms are based on Two-Stage detection, with poor performance. This paper proposes a pedestrian system One-Stage, and uses One-Stage-based YOLO-Tiny, YOLO YOLO-SPP to test analyze scene detecting pedestrians different sizes. The results show that have average confidence, YOLO-Tiny has fast speed, suitable for scenarios.
A human 3D avatar is one of the important elements in metaverse, and modeling effect directly affects people's visual experience. However, body has a complex topology diverse details, so it often expensive, time-consuming, laborious to build satisfactory model. Recent studies have proposed novel method, implicit neural representation, which continuous representation method can describe objects with arbitrary at resolution. Researchers applied obtained more excellent results than traditional...
The numerical integration of phase-field equations is a delicate task which needs to recover at the discrete level intrinsic properties solution such as energy dissipation and maximum principle.Although theory for classical phase field models well established, corresponding time-fractional still incomplete.In this article, we study certain nonlocal-in-time energies using first-order stabilized semi-implicit L1 scheme.In particular, will establish fractional law weighted law.The extension (2...
Rationale and ObjectivesEarly identification for hematoma expansion can help improve patient outcomes. Presently, there are many methods to predict expansion. This study compared a variety of models find model suitable clinical promotion.Materials MethodsNon-contrast head CT images data were collected from 203 patients diagnosed with hypertensive intracerebral hemorrhage. Radiomics features extracted all images, the dataset was randomly divided into training validation sets (7:3 ratio) after...
Chinese Medical Named Entity Recognition (CMNER) is the basic task of information processing and intelligent medical service in field. In order to make full use characters words text improve effect CMNER, this paper proposes a recognition model based on character-based BiLSTM-CRF, which integrates lexicon character features. Firstly, word sequence text, C-ExSoftword method proposed integrate into model. According similarity glyph morphologically related forms field medicine, takes...