This is a jupyter notebook series of 5 notebooks introducing beginner concepts of event selection, machine learning and their application in Higgs physics research using the Keras package for python specifically taylored to the applications of the CMS group of the Institute for Experimental Physics UHH. The beginning of this tutorial is very much based on the "Online Course: Machine Learning for Physicists 2021" which is a thorough lecture series by Florian Marquard. You can find the full recorded lectures, tutorials and slides here.
Keras is an open-source machine learning framework built on the foundation of Tensorflow 2. It is increadibly versitile and is chosen by many scientific organizations like CERN and NASA for its ease of use and applicability.
The series will cover everything from event selection in CMS and choosing suitable observables for training, what a neural network is and how to train them. The jupyter notebook environment is very convenient for quick scripting and trying out ideas where your code can be run on a cell by cell workflow, so you are encouraged to play around with the given code and edit it to your wants and needs!
Notebooks 1 - 3 will not dive into Keras just yet but will introduce fundamental concepts of cut-based event selection and the structure and nature of neural networks using native python and numpy objects only. This will allow us to develop an understanding of what is going on behind the scenes. Later on, many of the finicky details of neural networks will be managed by Keras such that we can focus on the physics problem at hand and can turn our full attention to the research to be conducted!
- numpy
- tensorflow
- keras
- matplotlib
- awkward
- uproot
- hist
- mplhep
- vector
- sklearn
The folder "images" contains all the images that are used in the notebooks. They images are taken from the internet and the source links are given at the bottom of the notebooks. The folder "misc" contains miscellaneous material related to the notebooks like output images or animations.
tbd