• nnetsauce's new version
  • Installing nnetsauce for Python
  • About nnetsauce for R

nnetsauce’s new version

A new version of nnetsauce, v0.12.0, is available on PyPI and for conda. It’s been mostly tested on Linux and macOS platforms. For Windows users: you can use the Windows Subsystem for Linux in case it doesn’t work directly on your computer.

As a reminder, nnetsauce does Statistical/Machine Learning (regression, classification, and time series forecasting for now) using randomized and quasi-randomized neural networks layers. More precisely, every model in nnetsauce is based on components g(XW + b), where:

  • X is a matrix containing explanatory variables and optional clustering information. Clustering the inputs helps in taking into account data’s heterogeneity before model fitting.
  • W creates new, additional explanatory variables from X. W can be drawn from various random and quasi-random sequences.
  • b is an optional bias parameter.
  • g is an activation function such as ReLU or the hyperbolic tangent, that makes the combination of explanatory variables – through W – nonlinear.

Examples of use of nnetsauce are available on GitHub, here (including R Markdown examples) and here.


v0.12.0 is an important release, because it’s totally written in Python (using numpy, scipy, jax, and scikit-learn), and doesn’t use C++ nor Cython anymore. Because of this, nnetsauce is faster to install, and easier to maintain.

If you like using nnetsauce, do not hesitate to star the repo or submit a pull request!

Installing nnetsauce for Python

  • 1st method: by using pip at the command line for the stable version
pip install nnetsauce
  • 2nd method: using conda (Linux and macOS only for now)
conda install -c conda-forge nnetsauce 
  • 3rd method: from Github, for the development version
pip install git+

or in a virtual environment:

git clone
cd nnetsauce
make install

About nnetsauce for R

The R version is discontinued. Well, ‘discontinued’ until I finally wrap my head around it… If you’re interested in solving this issue, and therefore, using nnetsauce for R, everything happens in this R script. You can submit a pull request (and star the repo ;) )!