deepregression

This is a refactored version of the old deepregression package.

Installation

To install the package, use the following command:

devtools::install_github("neural-structured-additive-learning/deepregression")

Note that the installation requires additional packages (see below) and their installation is currently forced by deepregression.

Requirements

The requirements are given in the DESCRIPTION. If you load the package manually using devtools::load_all, make sure the following packages are availabe:

• Matrix
• dplyr
• keras
• mgcv
• reticulate
• tensorflow
• tfprobability

If you set up a Python environment for the first time, install reticulate and run the check_and_install function from the deepregression package. This tries to install miniconda, TF 2.5.0, TFP 0.12 and keras 2.5.0, which seems to be the most reliable setup for deepregression at the moment.

Troubleshooting

Python Path and Conda / Virtual Environment

If R does not find Python or installed Python packages, check if the Python version and environment in R is set to the correct path. You can find your Python installations e.g. like this. To check if all the modules have been installed correctly -- just as a sanity check, because you can do the installation of modules also from inside R using reticulate::py_install -- you can use pip freeze other similar approaches. Finally, to check whether R also uses the Python version and environment you have installed all those modules into, you can force R to use a specific Python version using

reticulate::use_python("path/to/python/path", required = TRUE)

directly after starting your R session. Similar, you can force the usage of a virtual environment

reticulate::use_virtualenv("path/to/venv", required = TRUE)

or Conda environment

reticulate::use_condaenv("path/to/condaenv", required = TRUE)

again, directly after starting your R session.

How to cite this?

For the methodology, please cite the following preprint:

@article{rugamer2020unifying,
  title={Semi-Structured Deep Distributional Regression: Combining Structured Additive Models and Deep Learning},
  author={R{\"u}gamer, David and Kolb, Chris and Klein, Nadja},
  journal={arXiv preprint arXiv:2002.05777},
  year={2020}
}

For the software, please cite:

@article{rugamer2021deepregression,
  title={deepregression: a Flexible Neural Network Framework for Semi-Structured Deep Distributional Regression},
  author={David \textbf{R{\"u}gamer} and Chris Kolb and Cornelius Fritz and Florian Pfisterer and Philipp Kopper and Bernd Bischl and Ruolin Shen and Christina Bukas and Lisa Barros de Andrade e Sousa and Dominik Thalmeier and Philipp Baumann and Lucas Kook and Nadja Klein and Christian L. M{\"u}ller},
  year={2022},
  eprint={2104.02705},
  archivePrefix={arXiv},
  journal={Journal of Statistical Software},
  note={Provisionally Accepted}
}

How to use this?

See recent arXiv version

Python version

A Python version of the package is available here.

Related literature

The following works are based on the ideas implemented in this package:

• Original Semi-Structured Deep Distributional Regression Proposal
• Neural Mixture Distributional Regression
• Deep Conditional Transformation Models
• Semi-Structured Deep Piecewise Exponential Models
• Combining Graph Neural Networks and Spatio-temporal Disease Models to Predict COVID-19 Cases in Germany

People that contributed

Many thanks to following people for helpful comments, issues, suggestions for improvements and discussions:

• Andreas Bender
• Christina Bukas
• Oliver Duerr
• Patrick Kaiser
• Nadja Klein
• Philipp Kopper
• Christian Mueller
• Julian Raith
• Fabian Scheipl
• Matthias Schmid
• Max Schneider
• Ruolin Shen
• Almond Stoecker
• Dominik Thalmeier
• Kang Yang