cm-tpm 1.0.1

A Python package for data imputation using Continuous Mixtures of Tractable Probabilistic Models

cm-tpm (Continuous Mixtures of Tractable Probabilistic Models) is a Python package designed for efficient and scalable data imputation using probabilistic circuits. The package provides a flexible and user-friendly API, making it easy to integrate into data preprocessing pipelines. The package is distributed under the MIT licence.

The project was started in 2025 by Hakim Agni under the supervision of Thomas Krak at Eindhoven University of Technology. It implements the data imputation method described in the paper Continuous Mixtures of Tractable Probabilistic Models, by Alvaro Correia, Gennaro Gala, Erik Quaeghebeur, Cassio de Campos and Robert Peharz.

Installation

Dependencies

cm-tpm requires:

• Python (>= 3.10)
• NumPy (>= 1.22.4)
• PyTorch (>= 2.6.0)
• SciPy (>= 1.13.0)
• pandas (>= 2.2.2)

User installation

The easiest way to install cm-tpm is using pip:

pip install cm-tpm

Optional Dependencies

To install all optional dependencies:

pip install cm-tpm[all]

You can also install them individually:

• Excel file support (.xlsx), install openpyxl:

pip install cm-tpm[excel] 

• Parquet and Feather file support (.parquet. .feather), install pyarrow:

pip install cm-tpm[parquet]

• Progress bars (for verbose/debug mode), install tqdm:

pip install cm-tpm[tqdm]

Development

Source code

You can check the latest source code with the command:

git clone https://github.com/Hakim-Agni/cm-tpm.git

Install development dependencies:

pip install -r 'requirements.txt'

Testing

After installation, you can launch the test suite from outside the source directory (this requires pytest to be installed):

pytest tests

Help and Support

Documentation

The full documentation will be published soon.

Communication

If you have any questions or feedback, feel free to reach out via:

• GitHub Discussions: (coming soon).