In the following we explain how to use SMAC3. We explain the framework and the options you can use to configure SMAC.

SMAC stands for Sequential Model-Based Algorithm Configuration and uses Bayesian optimization to configure the hyperparameters of an algorithm. Instance features can be used to optimize the algorithm on a certain set of instances.

There are two ways to use SMAC: You can use the command line to optimize algorithms that are invoked via a bash command, but you can also use SMAC directly in Python.

We provide examples for the usage of SMAC in the Quickstart guide.

SMAC is written in Python 3 and hosted on GitHub.