machine-learning-tutorial / bayesian-optimization

An introduction to Bayesian optimization with an example of accelerator tuning task.

Tutorial on introduction to Bayesian optimization

• Download the repository
• Getting started
• Running the tutorial
• BO with scikit-learn

Material for this tutorial

• The tutorial in slide form can be found here

Download the repository

Get the repository with Git

You will need to have Git previously installed in your computer. To check if you have it installed, open your terminal and type:

git --version

Git installation in macOS

brew update
brew install git

Git installation in Linux

In Ubuntu/Debian

sudo apt install git

In CentOS

sudo yum install git

Once you have Git installed open your terminal, go to your desired directory, and type:

git clone https://github.com/machine-learning-tutorial/bayesian-optimization
cd bayesian-optimization

Get the repository with direct download

Open your terminal, go to your desired directory, and type:

wget https://github.com/machine-learning-tutorial/bayesian-optimization/archive/refs/heads/main.zip
unzip main.zip
cd bayesian-optimization

Getting started

You need to install the dependencies before running the notebooks.

Using conda

If you don't have conda installed already and want to use conda for environment management, you can install the miniconda as described here.

• Create a conda env with conda create -n bo-tutorial python=3.10
• Activate the environment with conda activate bo-tutorial
• Install the required packages via pip install -r requirements.txt.
• Run the following commands:

python -m jupyter contrib nbextension install --user
python -m jupyter nbextension enable varInspector/main

• After the tutorial you can remove your environment with conda remove -n bo-tutorial --all

Using venv only

If you do not have conda installed:

Alternatively, you can create the virtual env with venv in the standard library

python -m venv bo-tutorial

and activate the env with $ source /bin/activate (bash) or C:> /Scripts/activate.bat (Windows)

Then, install the packages with pip within the activated environment

python -m pip install -r requirements.txt
python -m jupyter contrib nbextension install --user
python -m jupyter nbextension enable varInspector/main

Afterwards, you should be able to run the provided notebooks.

Running the tutorial

After installing the package

You can start the jupyter notebook in the terminal, and it will start a browser automatically

python -m jupyter notebook

Alternatively, you can use supported Editor to run the jupyter notebooks, e.g. with VS Code.

Jupyter Notebooks

Use cmd+Enter to execute one cell block

BO with scikit-learn

In the sklearn-gp folder, there's an additional notebook explaining the BO concepts using only the scikit-learn package.

Citing the tutorial

This tutorial is registered Zenodo. Please use this DOI when citing this code:

@software{andrea_santamaria_garcia_2024_10723703,
  author       = {Santamaria Garcia, Andrea and
                  Xu, Chenran},
  title        = {{Tutorial on introduction to Bayesian optimization}},
  month        = {02},
  year         = {2024},
  publisher    = {Zenodo},
  version      = {v1.0},
  doi          = {10.5281/zenodo.10723703},
  url          = {https://doi.org/10.5281/zenodo.10723703}
}

Disclaimer

The content of this repository was developed by the AI4Accelerators team at the Institute of Beam Physics and Technology (IBPT), Karlsruhe Institute of Technology.