This repository serves as a curated hub for deep learning techniques applied to time-series imputation, targeting researchers and practitioners in the field. It provides a comprehensive list of essential papers, toolkits, and resources for handling missing data in time-series, aiming to streamline research and development in this area.

How It Works

The project acts as a meta-repository, aggregating and categorizing key resources for time-series imputation. It highlights the TSI-Bench paper and its associated code for benchmarking, alongside a collection of Python libraries like PyPOTS, TSDB, BenchPOTS, and PyGrinder for data loading, preprocessing, and imputation. The core value lies in its extensive, categorized list of academic papers and other repositories, offering a structured overview of the state-of-the-art and foundational work in the domain.

Quick Start & Requirements

• The primary interaction is through exploring the listed papers and toolkits.
• Specific toolkits (e.g., PyPOTS, TSDB) are Python-based and can be installed via pip.
• Requirements vary per toolkit but generally include Python and standard data science libraries.
• Links to official documentation and code are provided for individual toolkits and papers.

Highlighted Details

• Features the TSI-Bench paper and code for benchmarking time-series imputation methods.
• Curates a comprehensive list of over 50 influential papers on time-series imputation from 2016 to 2025, covering various deep learning architectures.
• Includes a collection of Python toolkits for data loading, preprocessing, and imputation, such as PyPOTS, TSDB, and BenchPOTS.
• Offers links to related resources, including articles on general missingness and imputation, and repositories focused on time-series transformers and foundation models.

Maintenance & Community

The repository is maintained by Wenjie Du and the PyPOTS Research community. Contributions for updating resources are welcomed. Citation information for related papers is provided.

Licensing & Compatibility

The repository itself does not appear to have a specific license listed. Individual toolkits and papers linked within will have their own licenses, which may vary. Compatibility for commercial use would depend on the licenses of the specific toolkits and datasets referenced.

Limitations & Caveats

This repository is primarily a curated list and does not offer a unified, runnable imputation framework itself. Users must interact with and install the individual toolkits mentioned. The "Awesome" nature implies a subjective selection of resources.