A5051382550- Time Series Analysis and Forecasting
- Neural Networks and Applications
- Stock Market Forecasting Methods
- Complex Systems and Time Series Analysis
- Data Stream Mining Techniques
- Music and Audio Processing
- Gaussian Processes and Bayesian Inference
- Advanced X-ray and CT Imaging
- Anomaly Detection Techniques and Applications
- Energy Load and Power Forecasting
- Water Quality and Pollution Assessment
- Advanced Neural Network Applications
- Anatomy and Medical Technology
- Water Treatment and Disinfection
- Model Reduction and Neural Networks
- Wastewater Treatment and Reuse
Huazhong University of Science and Technology
2022-2024
Huaqiao University
2023
North China Institute of Aerospace Engineering
2009
Recognition of surgical instruments is a key part the post-operative check and inspection instrument packaging. However, manual inventorying prone to counting errors. The achievement automated identification holds potential significantly mitigate occurrence medical accidents reduce labor costs. In this paper, an improved You Only Look Once version 5 (YOLOv5) algorithm proposed for recognition instruments. Firstly, squeeze-and-excitation (SE) attention module added backbone improve feature...
Reclaimed water is the valid approach to solve problem of resource scarcity and prevent secondary pollution. Based on case study Gaobeidian sewage treatment plant Beijing, security agricultural environment reclaimed irrigation was studied by statistic analysis quality experiment. With Sewage Treatment Plant qualities rules change were studied. The method principal component (PCA) used extract main indices that can delegate quality. It proved coliform group, total salt, N, LAS heavy metal...
Time series classification faces two unavoidable problems. One is partial feature information and the other poor label quality, which may affect model performance. To address above issues, we create a correction method to time data with meta-learning under multi-task framework. There are three main contributions. First, train two-branch neural network in outer loop. While model-agnostic inner loop, use pre-existing models way jointly update meta-knowledge so as help us achieve adaptive...
Time series classification faces two unavoidable problems. One is partial feature information and the other poor label quality, which may affect model performance. To address above issues, we create a correction method to time data with meta-learning under multi-task framework. There are three main contributions. First, train two-branch neural network for outer loop. While in modelagnostic inner loop, use pre-existing models way jointly update meta-knowledge, makes us achieve adaptive...
In this article, we employ a collection of stochastic differential equations with drift and diffusion coefficients approximated by neural networks to predict the trend chaotic time series which has big jump properties. Our contributions are, first, propose model called L\'evy induced equation network, explores compounded $\alpha$-stable motion complex data solve problem through network approximation. Second, theoretically prove that numerical solution our algorithm converges in probability...