SMAC is a tool for algorithm configuration to optimize the parameters of arbitrary algorithms, including hyperparameter optimization of Machine Learning algorithms. The main core consists of Bayesian Optimization in combination with an aggressive racing mechanism to efficiently decide which of two configurations performs better.
SMAC3 is written in Python3 and continuously tested with Python 3.7, 3.8 and 3.9. Its Random Forest is written in C++. In further texts, SMAC is representatively mentioned for SMAC3.
Create a new environment with python 3.9 and make sure swig is installed either on your system or inside the environment. We demonstrate the installation via anaconda in the following:
Create and activate environment:
conda create -n SMAC python=3.9
conda activate SMAC
Install swig:
conda install gxx_linux-64 gcc_linux-64 swig
Install SMAC via PyPI:
pip install smac
Or alternatively, clone the environment:
git clone https://github.com/automl/SMAC3.git && cd SMAC3
pip install -r requirements.txt
pip install .
We refer to the documention for further installation options.
import numpy as np
from sklearn.ensemble import RandomForestClassifier
from ConfigSpace import ConfigurationSpace
from ConfigSpace.hyperparameters import UniformIntegerHyperparameter
from smac.facade.smac_bb_facade import SMAC4BB
from smac.scenario.scenario import Scenario
X_train, y_train = np.random.randint(2, size=(20, 2)), np.random.randint(2, size=20)
X_val, y_val = np.random.randint(2, size=(5, 2)), np.random.randint(2, size=5)
def train_random_forest(config):
"""
Trains a random forest on the given hyperparameters, defined by config, and returns the accuracy
on the validation data.
Input:
config (Configuration): Configuration object derived from ConfigurationSpace.
Return:
cost (float): Performance measure on the validation data.
"""
model = RandomForestClassifier(max_depth=config["depth"])
model.fit(X_train, y_train)
# define the evaluation metric as return
return 1 - model.score(X_val, y_val)
if __name__ == "__main__":
# Define your hyperparameters
configspace = ConfigurationSpace()
configspace.add_hyperparameter(UniformIntegerHyperparameter("depth", 2, 100))
# Provide meta data for the optimization
scenario = Scenario({
"run_obj": "quality", # Optimize quality (alternatively runtime)
"runcount-limit": 10, # Max number of function evaluations (the more the better)
"cs": configspace,
})
smac = SMAC4BB(scenario=scenario, tae_runner=train_random_forest)
best_found_config = smac.optimize()More examples can be found in the documention.
This program is free software: you can redistribute it and/or modify it under the terms of the 3-clause BSD license (please see the LICENSE file).
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.
You should have received a copy of the 3-clause BSD license along with this program (see LICENSE file). If not, see here.
SMAC3 is developed by the AutoML Groups of the Universities of Hannover and Freiburg.
If you have found a bug, please report to issues. Moreover, we are appreciating any kind of help. Find our guidlines for contributing to this package here.
If you use SMAC in one of your research projects, please cite us:
@misc{lindauer2021smac3,
title={SMAC3: A Versatile Bayesian Optimization Package for Hyperparameter Optimization},
author={Marius Lindauer and Katharina Eggensperger and Matthias Feurer and André Biedenkapp and Difan Deng and Carolin Benjamins and René Sass and Frank Hutter},
year={2021},
eprint={2109.09831},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
Copyright (C) 2016-2021 AutoML Group.