Python package for conditional density estimation. It either wraps or implements diverse conditional density estimators.
This package provides pass-through access to all the functionalities of nflows.
pyknos requires Python 3.8 or higher. A GPU is not required, but can lead to speed-up
in some cases. We recommend using a
conda virtual environment
(Miniconda installation instructions).
If conda is installed on the system, an environment for installing pyknos can be
created as follows:
$ conda create -n pyknos_env python=3.12 && conda activate pyknos_env
To install pyknos from PyPI run
python -m pip install pyknos
To install and add pyknos to a project with pixi, from the project directory run
pixi add pyknos
and to install into a particular conda environment with conda, in the activated environment run
conda install --channel conda-forge pyknos
See the sbi repository for examples of using pyknos.
pyknós (πυκνός) is the transliterated Greek root for density (pyknótita) and also means sagacious.
This program is free software: you can redistribute it and/or modify it under the terms of the Apache License 2.0., see LICENSE for more details.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
Thanks to Artur Bekasov, Conor Durkan and George Papamarkarios for their work on nflows.
The MDN implementation in this package is based on Conor M. Durkan's.