- Electronic Packaging and Soldering Technologies
- 3D IC and TSV technologies
- Advanced Welding Techniques Analysis
- Non-Destructive Testing Techniques
- Advanced Vision and Imaging
- Adhesion, Friction, and Surface Interactions
- Domain Adaptation and Few-Shot Learning
- Advanced Neural Network Applications
- Advanced machining processes and optimization
- Ultrasonics and Acoustic Wave Propagation
- Sparse and Compressive Sensing Techniques
- Optical measurement and interference techniques
- Industrial Vision Systems and Defect Detection
- Advanced MEMS and NEMS Technologies
- Machine Learning and ELM
- Mechanical Behavior of Composites
- Manufacturing Process and Optimization
- Robotics and Sensor-Based Localization
- Advanced Image and Video Retrieval Techniques
- Image Processing Techniques and Applications
- Face and Expression Recognition
- Vehicle License Plate Recognition
- Remote Sensing and LiDAR Applications
- Advanced Measurement and Detection Methods
- Advanced Measurement and Metrology Techniques
Nanjing Surveying and Mapping Research Institute (China)
2025
Nanjing Institute of Industry Technology
2023-2024
Nanjing Institute of Technology
2007-2024
Dalian University of Technology
2024
Central South University
2008-2023
Anhui University
2008-2023
Lanzhou Jiaotong University
2023
Chang'an University
2022
Changsha Normal University
2020
Hunan Normal University
2020
In multi-task learning (MTL), multiple related tasks are learned jointly by sharing information across them. Many MTL algorithms have been proposed to learn the underlying task groups. However, those methods limited groups at only a single level, which may be not sufficient model complex structure among in many real-world applications. this paper, we propose Multi-Level Task Grouping (MeTaG) method multi-level grouping instead of one level tasks. Specifically, assuming number levels H,...
In multi-task learning (MTL), multiple related tasks are learned jointly by sharing information according to task relations. One promising approach is utilize the given tree structure, which describes hierarchical relations among tasks, learn model parameters under regularization framework. However, such a priori rarely available in most applications. To best of our knowledge, there no work structure and simultaneously framework this paper, we develop TAsk Tree (TAT) for MTL achieve this. By...
Multi-task learning (MTL) seeks to improve the generalization performance by sharing information among multiple tasks. Many existing MTL approaches aim learn low-rank structure on weight matrix, which stores model parameters of all tasks, achieve task sharing, and as a consequence trace norm regularization is widely used in literature. A major limitation these based that singular values matrix are penalized simultaneously, leading impaired estimation recovering larger matrix. To address...
Multi-task learning seeks to improve the generalization performance by sharing common information among multiple related tasks. A key assumption in most MTL algorithms is that all tasks are related, which, however, may not hold many real-world applications. Existing techniques, which attempt address this issue, aim identify groups of using group sparsity. In paper, we propose a probabilistic tree sparsity (PTS) model utilize structure obtain sparse solution instead structure. Specifically,...
The features of ultrasonic bonding interface were inspected by using a high resolution transmission electron microscope. Stress was analysed the finite elements simulation. Results show that stress caused vibration, which increased dislocation density inside metal crystalline lattice provides fast diffusion channels, and provided driving force for atom inter-diffusion. 'Short-circuit diffusion' during is more prominent than crystal diffusion. For given parameters, depth at Au/Al about...
Current model quantization methods have shown their promising capability in reducing storage space and computation complexity. However, due to the diversity of forms supported by different hardware, one limitation existing solutions is that usually require repeated optimization for scenarios. How construct a with flexible has been less studied. In this paper, we explore one-shot network regime, named Elastic Quantization Neural Networks (EQ-Net), which aims train robust weight-sharing...
Thermosonic flip chip bonding is used in certain fine pitch IC packaging for its unique features. By using this process paper, 1 mm times with 8 gold bumps has been bonded onto a silver-coated substrate. Dynamical properties of transducer system, which the key component providing ultrasonic energy, have investigated finite element model (FEM) and measurement impedance analyzer laser Doppler vibrometer (LDV). The simulation results show that vibrates by coupling all excited modes, therefore...
Thermosonic flip chip (TSFC) bonding is a developing area in array microelectronic interconnection technology, and ultrasonic vibration plays an important role TSFC bonding. To understand the effects at interface, lab bonder constructed, realized, some (plastic deformation of bumps on bond pads, atom diffusion, increased dislocation density) are observed. A dynamic finite element model developed to analyze stress strain distribution interface. The results our study show that causes large...
Digital image processing technology presents new challenges for the security of information, which makes authentication as a problem to be addressed. watermarking plays an important role in multimedia information security. This paper proposed dual method do tamper location and self-recovery, utilizes self-embedding hide eigenface into low frequence sub band LL restore region. Experimental results show availability algorithm.
The effect of the temperature on bondability and bonding process for wire are investigated. Bondability is characterized by shear strength represented input output power ultrasonic transducer. A laser Doppler vibrometer Labview software were used to record velocity, voltage current transducer at different settings. K-type thermocouple sensor was measure temperature. Experimental results show that unsuccessful happens low temperature, over appears if too high. Only when appropriate settings,...
In thermosonic wire bonding, a free air ball (FAB) created by electronic flame-off (EFO) is crucial to reliable first bond. To reveal the mechanisms of FAB and ripple formation, dynamic formation process was monitored analyzed with high-speed camera digital data acquisition system. found occur in four stages: initiation, preheating, melting, solidification. The speed model depend only on EFO current. can be as high 90.4 μm/ms at an current 60 mA. A minimum time required for which depends...
In order to investigate the performance of microprobe in integrated circuit (IC) testing, a novel testing system with four-wire cable tester and contact force sensor is deployed using high-density microprobes whose resistance are automatically measured saved LabVIEW. The statistically analyzed for DE1-030EB40-010 type probes adjusted parameters. results show that displacement effect significant testing. addition, not stable small value until exceeds 300 μm. Furthermore, velocity varying from...
A testing circuit of probe is designed to understand the real-time characteristics during process. The constant resistance voltage and are measured simultaneously by DSOX2022A oscilloscope, calculated. result shows that varies from initial infinity approximately $1~\Omega $ at 0.2 ms when in contact with copper substrate. gradually reduced relatively stable value increasing compressed displacement. increases badly 30 times angle between changes 0° 2°, reduces a after 20-30 testing.