- VLSI and Analog Circuit Testing
- Integrated Circuits and Semiconductor Failure Analysis
- Engineering and Test Systems
- Advanced Neural Network Applications
- Advanced Multi-Objective Optimization Algorithms
- VLSI and FPGA Design Techniques
- Advancements in Semiconductor Devices and Circuit Design
- Adversarial Robustness in Machine Learning
- Domain Adaptation and Few-Shot Learning
- Anomaly Detection Techniques and Applications
- Machine Learning and ELM
- Software Testing and Debugging Techniques
- Industrial Vision Systems and Defect Detection
- Metaheuristic Optimization Algorithms Research
- Smart Agriculture and AI
- Spectroscopy and Chemometric Analyses
- Fault Detection and Control Systems
- Advanced Image and Video Retrieval Techniques
- Probabilistic and Robust Engineering Design
- Hydraulic and Pneumatic Systems
- Evolutionary Algorithms and Applications
- Advanced DC-DC Converters
- Microstructure and mechanical properties
- Energy Load and Power Forecasting
- Piezoelectric Actuators and Control
Jilin Agricultural University
2023-2025
Harbin Institute of Technology
2019-2024
Central South University
2023-2024
General Hospital of Guangzhou Military Command
2024
Zhengzhou University
2024
Dongbei University of Finance and Economics
2024
University of Electronic Science and Technology of China
2010-2023
Northeastern University
2015-2023
Johns Hopkins University
2018-2022
Southwest University of Science and Technology
2020
Optimizing a deep neural network is fundamental task in computer vision, yet direct training methods often suffer from over-fitting. Teacher-student optimization aims at providing complementary cues model trained previously, but these approaches are considerably slow due to the pipeline of few generations sequence, i.e., time complexity increased by several times. This paper presents snapshot distillation (SD), first framework which enables teacher-student one generation. The idea SD very...
Despite the impressive representation capacity of vision transformer models, current light-weight models still suffer from inconsistent and incorrect dense predictions at local regions. We suspect that power their self-attention mechanism is limited in shallower thinner networks. propose Lite Vision Transformer (LVT), a novel network with two enhanced mechanisms to improve model performances for mobile deployment. For low-level features, we introduce Convolutional Self-Attention (CSA)....
We focus on the problem of training a deep neural network in generations. The flowchart is that, order to optimize target (student), another (teacher) with same architecture first trained, and used provide part supervision signals next stage. While this strategy leads higher accuracy, many aspects (e.g., why teacher-student optimization helps) still need further explorations.This paper studies from perspective controlling strictness teacher network. Existing approaches mostly hard...
We focus on the problem of training a deep neural network in generations. The flowchart is that, order to optimize target (student), another (teacher) with same architecture first trained, and used provide part supervision signals next stage. While this strategy leads higher accuracy, many aspects (e.g., why teacher-student optimization helps) still need further explorations. This paper studies from perspective controlling strictness teacher network. Existing approaches mostly hard...
Soft-fault diagnosis and tolerance are two challenging problems in analog-circuit fault diagnosis. Although many analog faults can be diagnosed theoretically, they cannot accurately due to the influence of component tolerance. This paper proposes approaches tolerance-handling method soft-fault First, slope model its theoretical proof presented. In linear circuits, voltage equation between nodes is expressed by a point–slope-form which point determined nominal values on selected nodes, slope,...
Image tokenizers form the foundation of modern text-to-image generative models but are notoriously difficult to train. Furthermore, most existing rely on large-scale, high-quality private datasets, making them challenging replicate. In this work, we introduce Text-Aware Transformer-based 1-Dimensional Tokenizer (TA-TiTok), an efficient and powerful image tokenizer that can utilize either discrete or continuous 1-dimensional tokens. TA-TiTok uniquely integrates textual information during...
Rice is an important part of the food supply, its different varieties in terms quality, flavor, nutritional value, and other aspects differences, directly affect subsequent yield economic benefits. However, traditional rice identification methods are time-consuming, inefficient, prone to damage. For this reason, study proposes a deep learning-based method classify identify with flavors fast non-destructive way. In experiment, 19 categories japonica seeds were selected, total 36735 images...
This paper introduces the COCONut-PanCap dataset, created to enhance panoptic segmentation and grounded image captioning. Building upon COCO dataset with advanced COCONut masks, this aims overcome limitations in existing image-text datasets that often lack detailed, scene-comprehensive descriptions. The incorporates fine-grained, region-level captions ensuring consistency improving detail of generated captions. Through human-edited, densely annotated descriptions, supports improved training...
Objective Although preoperative prediction of axillary lymph nodes status has been achieved using radiomics and combined models, there is a dearth research on internal mammary node (IMN) metastasis prediction. We developed predictive model by combining clinicopathological factors with dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) to accurately predict IMN in breast cancer. Methods Patients who had no evidence images but underwent sentinel biopsy (IM-SLNB) were included this...
Endodontic treatment is performed to treat inflamed or infected root canal system of any involved teeth. It estimated that 22.3 million endodontic procedures are annually in the USA. Preparing a proper access cavity before cleaning/shaping (instrumentation) among most important steps achieve successful outcome. However, accidents such as perforation, gouging, ledge and transportation may occur during procedure because an improper incomplete design. To reduce prevent these errors treatments,...
The near-optimal test-point set selection for an analog fault dictionary is formulated as a heuristic depth-first graph search problem. Then, the test point process becomes graph-node-expanding process. During of expansion, information-theoretic concepts entropy are used to develop criterion how choose intermediate node expand. If has already isolated those faults that hard isolate, then residual can easily be isolated. difficulty isolating given evaluated by concept entropy. harder...
Owing to the lack of feasible fault modeling method, hard (open and short) faults, discretized parametric faults are still mostly used models. These models cannot characterize all soft (parameter shifting) because that parameter analog element be continuity character. To address this concern, a complex field method is presented first. If happens passive xi in linear circuit, real (U <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">r</sub> )...
The insulated gate bipolar transistor (IGBT) is a kind of excellent performance switching device used widely in power electronic systems. How to estimate the remaining useful life (RUL) an IGBT ensure safety and reliability electronics system currently challenging issue field reliability. aim this paper develop prognostic technique for estimating IGBTs’ RUL. There need efficient algorithm that able support in-situ decision-making. In paper, novel prediction model with complete structure...
Lithium-ion batteries (LiBs) are the most important part of electric vehicle (EV) systems. Because there two different degradation rates during LiB degradation, many two-phase models for LiBs. However, these methods do not consider randomness changing point in model and cannot update change time real time. Therefore, this paper proposes a method based on combination Wiener an extreme learning machine (ELM). The is used to derive mathematical expression remaining useful life (RUL), ELM...
Hard (open and short) faults discrete parameter (DPFs) are the mostly used fault models in simulation-before-test (SBT) method. Because of analog element is continuous, DPF cannot elaborately characterize all possible continuous (CPF) occurring circuit. To address such a problem, genetic algorithm (GA)-based simulation after test (SAT) diagnosis method proposed this paper. The transformed into an optimization problem. genes represent values potential faulty components. circuit response...
Hard (open and short) faults discrete parameter (DPF) are the mostly used fault models in simulation-before-test (SBT) method. Because that of analog element is continuous, DPF can not elaborately characterize all possible continuous (CPF) occurring circuit, let alone double soft fault. To address such problem, a genetic algorithm (GA) based simulation after test (SAT) diagnosis method proposed this paper. The transformed into an optimization problem. genes represent values potential faulty...
The modeling approaches of power converters occupy an important position in electronic systems and have made considerable progress over the past years. Continuous linearization techniques are reviewed, including state-space average model, generalized averaged small-signal describing function method. A Buck converter with PWM modulation voltage-mode control is taken as example to compare advantages disadvantages different methods through simulation analysis. Moreover, corresponding equivalent...
Rice is one of the most important crops for food supply, and there are multiple differences in quality rice different geographic regions, which have a significant impact on subsequent yields economic benefits. The traditional identification methods time-consuming, inefficient, delicate. This study proposes deep learning-based method fast non-destructive classification grown environments. experiment collected with name Ji-Japonica 830 from 10 total 10,600 grains were obtained, fronts...
This paper proposes a method for recognizing the origin of Saposhnikovia divaricata using IResNet model to achieve computer vision-based classification. Firstly, we created small sample dataset and applied data augmentation techniques enhance its diversity. After that, introduced hierarchical residual connection block in early stage original expand perceptual field neural network extraction scale features. Meanwhile, depth-separable convolution operation was adopted later replace...
Once all available measurements are determined, the highest testability index of a complex system is determined. To achieve such with lowest test cost, AND/OR graph search algorithms were developed for years to determine an optimal or near-optimal sequence. However, in most cases, achieving induces extremely high cost. The purpose this paper optimize set and sequence so as cut down cost while keeping required, not necessarily highest, FIR (Fault Isolation Rate) satisfied. Traditionally,...