Laurits7 / HyperEvolution

Algorithms for hyperparameter optimization

HyperEvolution

Algorithms for hyperparameter optimization

Installation

First make sure you are using python3 python --version. To set python3 as your default, an easy way is just to alias python to python3:

echo 'alias python="python3"' >> $HOME/.bashrc
echo 'alias pip="pip3"' >> $HOME/.bashrc
source $HOME/.bashrc

Clone the repository:

git clone [e-mail]:Laurits7/HyperEvolution.git
cd HyperEvolution

Create a virtual environment and activate it:

python -m venv Hopt
source Hopt/bin/activate

And install the package (-e for the editable version):

pip install -e .

Now every time you need to run the code, you only need to source the environment again

Examples

The example optimizations can be found under examples

Additional notes:

The optimization algorithms presented here are doing function minimization, so in case you want to maximize (for example AUC) you need to return the score by the scoring function with a minus sign.

---

References:

The work presented here is based on these two papers:

Tani, Laurits, Diana Rand, Christian Veelken, and Mario Kadastik. 2021. “Evolutionary Algorithms for Hyperparameter Optimization in Machine Learning for Application in High Energy Physics.” The European Physical Journal C 81 (2): 1–9.

Tani, Laurits, and Christian Veelken. 2022. “Comparison of Bayesian and Particle Swarm Algorithms for Hyperparameter Optimisation in Machine Learning Applications in High Energy Physics.” arXiv Preprint arXiv:2201.06809.