A5080890470- Metal and Thin Film Mechanics
- Target Tracking and Data Fusion in Sensor Networks
- Integrated Circuits and Semiconductor Failure Analysis
- Corrosion Behavior and Inhibition
- Underwater Vehicles and Communication Systems
- Educational Technology and Assessment
- Silicon and Solar Cell Technologies
- Indoor and Outdoor Localization Technologies
- Industrial Vision Systems and Defect Detection
- Adhesion, Friction, and Surface Interactions
National Yang Ming Chiao Tung University
2023-2024
Knowledge Distillation (KD) is the procedure of extracting useful information from a previously trained model using an algorithm. The successful distillation pulls up distilling accuracy. In context compression, teacher provides softened labels that facilitate distillation. While some effort has been devoted to progressive training, and majority in literature teacher-student configuration or methodologies, less spent non-progressive meta-learning controlled distillation, training with...
Promoted model architectures or algorithms are crucial for intelligent manufacturing since developing them takes a lot of trial and error to embed the domain knowledge into models correctly. Especially in semiconductor manufacturing, whole processes depend on complicated physical equations sophisticated fine-tuning. Therefore, we use neuroevolution-based search optimized architecture automatically. The collector current value at particular bias silicon-germanium (SiGe) heterojunction bipolar...
In this paper, we propose a time-difference-of-arrival (TDOA) based outdoor localization approach using deep neural network (DNN) with novel weighted mean squared error (WMSE) loss. The weight is determined aided by the Cramer-Rao lower bound (CRLB) of accuracy. A training set rebalancing strategy proposed to approximate WMSE loss classical MSE and make process efficient, without need collecting more samples. Extensive numerical investigations ideal nonideal scenarios are included examine...