Machine learning tutorial

Pietro Vischia (pietro.vischia@cern.ch)

Originally developed as a tutorial held at ICNFP 2022 (Kolumpari, Crete).

Tutorial organization

Ideally you would be running the tutorial on your laptop, following the instructions and explanations given by me in the big screen in the room. If, for any reason, you cannot run the tutorial, you are welcome to just watch the tutorial steps being executed in the big screen by me.

How to run the tutorial on your local machine

1. Check out the code

git clone git@github.com:vischia/machine_learning_tutorial.git
cd machine_learning_tutorial/

or

git clone https://github.com/vischia/machine_learning_tutorial.git
cd machine_learning_tutorial/

2. Create a python environment and install requirements (follow one of the options 2.1, 2.2, or 2.3)

2.1 Using virtualenv

virtualenv -p python3.9 venv_tutorial
source venv_tutorial/bin/activate
pip install -r requirements.txt # or requirements_macos.txt on MacOS Monterey

A participant (Geoffrey Mullier) reports that on MacOS 12.5 virtualenv doesn't work, and that in that case python3 -m venv venv_tutorial works as intended.

To deactivate the environment, you should run deactivate from the command prompt.

2.2 Using conda

conda create --name venv_tutorial python==3.9.13
conda activate venv_tutorial
pip install -r requirements.txt # or requirements_macos.txt on MacOS Monterey

To deactivate the environment, you should run conda deactivate from the command prompt.

2.3 Using Google Colab (google account needed)

Go to Google Colab, select GitHub as a source, and fill in the path to this repository (https://github.com/vischia/machine_learning_tutorial). Possibly Google will ask for access to your GitHub account, although installing from a public third party repository should not require that, in principle.

When the colab instance is active, open the jupyter notebook train_hyp.ipynb and run the cell labelled "If you are using COLAB"

3. Run the tutorial

For local environments, run

jupyter notebook

and open train_hyp.ipynb in the browser window that is opened.

From Colab, open train_hyp.ipynb.