A5101761835- Sensor Technology and Measurement Systems
- Optical Systems and Laser Technology
- Industrial Vision Systems and Defect Detection
- Infrared Thermography in Medicine
- Cognitive Science and Mapping
- Transportation Planning and Optimization
- Smart Grid and Power Systems
- Traffic control and management
- Design Education and Practice
- Traffic Prediction and Management Techniques
- Adversarial Robustness in Machine Learning
- Evaluation Methods in Various Fields
- Power Systems and Renewable Energy
- Traditional Chinese Medicine Studies
- Advanced Neural Network Applications
Nanjing Normal University
2024-2025
Suzhou Institute of Biomedical Engineering and Technology
2025
Chinese Academy of Sciences
2025
Beijing Jiaotong University
2024
Large language models (LLMs) face inherent performance bottlenecks under parameter constraints, particularly in processing critical tokens that demand complex reasoning. Empirical analysis reveals challenging induce abrupt gradient spikes across layers, exposing architectural stress points standard Transformers. Building on this insight, we propose Inner Thinking Transformer (ITT), which reimagines layer computations as implicit thinking steps. ITT dynamically allocates computation through...
In the clinical stage, suspected hemolytic plasma may cause hemolysis illness, manifesting as symptoms such heart failure, severe anemia, etc. Applying a deep learning method to images significantly improves recognition accuracy, so that this paper proposes quality detection model based on improved "You Only Look Once" 5th version (YOLOv5). Then presented in and evaluation system were introduced into datasets, the average accuracy of final classification reached 98.7%. The results...
Traffic flow prediction can provide important reference data for managers to maintain traffic order, and also be based on personal travel plans optimal route selection. On account of the development sensors collection technology, large-scale road network historical effectively used, but their high non-linearity makes it meaningful establish effective models. In this regard, paper proposes a dual-stream cross AGFormer-GPT with prompt engineering prediction, which integrates occupancy speed as...
Transformer models encounter challenges in scaling hidden dimensions efficiently, as uniformly increasing them inflates computational and memory costs while failing to emphasize the most relevant features for each token. For further understanding, we study dimension sparsity observe that trained Transformers utilize only a small fraction of token dimensions, revealing an "activation flow" pattern. Notably, there are shared sub-dimensions with sustained activation across multiple consecutive...
China's economy is experiencing a real situation of slowing growth, this has major impact on the development electricity demand. It urgent to ask power trading center have an in-depth understanding and changes in purchase sale resources. Through research analysis domestic international system indicator system. Starting from these resource characteristics, paper comprehensively sorts summarizes sales resources, proposes for purchasing selling resources that suits business, providing...