Experiments accompanying Auto-sklearn 2.0: The Next Generation

by Matthias Feurer, Katharina Eggensperger, Stefan Falkner, Marius Lindauer and Frank Hutter

currently under submission - arXiv to appear

This repository is provided as-is and we do not provide any maintenance.
Necessary scripts will be ported to the main Auto-sklearn repository.

Code organization

This repository contains experiment scripts for the above mentioned paper submission. We
will clean up the code over time to make it easier to understand and reproduce our results
while keeping its functionality. The code is organized into two directories:

• experiment_scripts contains all scripts we used for the experiments. A manual
  describing in which order to execute the scripts is given in HowTo.md.
• notebooks contains several jupyter-notebooks which we used to analyze the results
  and produce the tables and plot given in the paper.

Installation and requirements

• Installation instructions and requirements will be provided soon.
• We are working on a cleaner version which does not require setting directories manually.

Further notes

Running the code for the 1 hour setting will cost roughly one to two CPU years and should be done on a high-performance compute cluster.