More about HKUST
A Survey on Deep Learning Models for Time Series Analysis
PhD Qualifying Examination Title: "A Survey on Deep Learning Models for Time Series Analysis" by Mr. Shuhan ZHONG Abstract: Time series analysis, a long-standing area of research, involves processing sequential data points collected over time to extract meaningful information for tasks such as forecasting, classification, imputation, and anomaly detection. Unlike image or text data, time series data is characterized by its complex temporal variations, posing unique challenges in identifying the underlying patterns. Classical time series methods have been instrumental in modeling these patterns but struggle with scalability and expressiveness in the face of large multivariate time series datasets. To this end, much effort has been made to apply deep learning methods for time series analysis, leading to a surge in related publications, which makes it challenging to keep pace with the developments in this field. This survey aims to provide a comprehensive overview of deep learning models for time series analysis across different tasks. Specifically, we review recent publications in this area and provide a taxonomy that groups diverse and relevant papers into categories by how they integrate classical time series methods into deep learning models, and how they innovate existing deep architectures. We present the underlying principles and discuss representative works. Besides, we summarize the most widely-used benchmarking datasets and protocols for different tasks with a quantitative evaluation of representative models. We then highlight key challenges and promising future research directions in this area. Date: Wednesday, 20 November 2024 Time: 4:00pm - 6:00pm Venue: Room 5501 Lifts 25/26 Committee Members: Prof. Gary Chan (Supervisor) Prof. Qiong Luo (Chairperson) Prof. Andrew Horner Prof. Xiaofang Zhou