The article discusses the release of the first open-source family of time series foundation models, contributing significantly to the development and evaluation processes. It introduces the 'Time Series Pile', a comprehensive collection of public time series data, used to pre-train transformer models addressing specific challenges in time series analysis. The benchmark evaluations reveal strong performance by MOMENT, especially in anomaly detection and classification tasks with limited datasets. Future work aims at applying MOMENT in real-world scenarios and exploring advanced techniques in multi-modal time series modeling.
"We release the first open-source family of time series foundation models and make contributions at all stages of the development and evaluation process."
"We first compile a large and diverse collection of public time series, called the Time Series Pile, and demonstrate its efficacy by pre-training high-performing time series foundation models from scratch."
Collection
[
|
...
]