PyPOTS: End-to-End Learning for Partially-Observed Time Series
A recently released tutorial presents PyPOTS, an open-source Python framework designed for comprehensive data mining and machine learning on partially-observed time series (POTS). It addresses various aspects such as simulating missing data, preprocessing, and evaluating models across tasks like imputation, forecasting, classification, clustering, and anomaly detection. The first part emphasizes practical applications using unified APIs and benchmark experiments, while the second part is aimed at developers and researchers, discussing custom models, domain-specific constraints, and engineering methodologies. This work can be found on arXiv under the identifier 2604.24041.
Key facts
- PyPOTS is an open-source Python ecosystem for partially-observed time series.
- The tutorial covers missingness simulation, preprocessing, training, and evaluation.
- Tasks include imputation, forecasting, classification, clustering, and anomaly detection.
- Part I is for practitioners with unified APIs and benchmarks.
- Part II is for developers and researchers on custom models and constraints.
- The tutorial is published on arXiv with ID 2604.24041.
- It addresses the problem of separating missing-value handling from downstream learning.
- The goal is to improve reproducibility and overall performance.
Entities
Institutions
- arXiv