Project Name
Development of Advanced Time Series Machine Learning Core Techniques and Integrated Tools
Project Goal
This project aims to develop breakthrough techniques and tools for time-series machine learning, including early prediction with multi-objective optimization, unsupervised learning, multi-task learning, and automated machine learning. The developed techniques can be applied to various domains and applications, particularly in anomaly prediction in the medical and environmental sensing fields.
Project Description
1. Time Series Early Prediction
＊Snippet representation learning
＊Early classification timing learning
2. Unsupervised Time Series Learning
＊Time series sampling method
＊Time series ensemble method
3. Temporal-based Deep Multi-task Learning
＊Auxiliary task generation
＊Dynamic task filtering
4. Integrated and Automated Time Series Learning Tools
＊High dimensional time series processing
＊Irregular time series processing
