This is a set of tutorials for the Machine Learning Hands-on Advanced Tutorial Session (HATS). They are intended to show you how to build machine learning models in python (Keras/TensorFlow) and use them in your ROOT-based analyses. We will build event-level classifiers for differentiating VBF Higgs and standard model background 4 muon events and jet-level classifiers for differentiating boosted W boson jets from QCD jets.
a-dataset-and-plot.ipynb: reading/writing datasets fromROOTfiles withuprootand plotting withmatplotlibb-dense.ipynb: building, training, and evaluating a fully connected (dense) neural network inKerasb.1-dense-pytorch.ipynb: building, training, and evaluating a fully connected (dense) neural network inPyTorchb.2-dense-bayesian-optimization.ipynb: building, training, and evaluating a fully connected (dense) neural network inKerasand exploring hyperparametersc-conv2d.ipynb: preprocessing, building, training, and evaluating a 2D convolutional neural network inKeras
There are multiple ways to run through this tutorial with varying levels of setup required.
In this setup, environment and data files have already been set up for you on available hardware. At Fermilab, to access, use the link: https://analytics-hub.fnal.gov:8080/ and sign in using your services account. You need to be granteed the cluster permissions first.
We will be setting up the environment using Miniconda wity Python3. This is wrapped in a Docker container for easy deployment.
# currently it's setup for Mac, change the file if you want to run on Linux
source install_miniconda3.sh
# setup conda environmeent and install needed packages
source setup.sh
Run the tutorial directly in a Docker container on your local machine (requires Docker installation)
docker run -p 127.0.0.1:8888:8888 mapsacosta/ml-hats:latest
You can launch notebooks in Binder for quick tests, but note this is not for resource-intensive computing:
The accompanying lecture is here