Classifying events in a nuclear physics experiment - an example

Standalone repository for classification example on simulated data that can be used in ML course.

Content

/notebooks

data_import_exploration

This notebook contains a walkthrough from the first import of the data, formatting the data, initial analysis and exploration, and finally storing the data in a desired format.

data_import_exploration_traces

A starting point for working with the traces datafile.

logistic

Scikit-Learn example on using logistic regression to classify the data. Presents the metrics we use to assess the performance of a binary classifier.

dense

Keras example on using a fully-connected neural network to classify the data

convolutional

Keras example on using a convolutional neural network to classify the data

helper_functions.py

Collection of functions developed throughout the notebooks that are useful to store for easier re-use.

/data

• CeBr10k_1.txt
  • file with 10000 events
• CeBr200k_Mix.txt.gz
  • file with 200k events, compressed to comply with githubs max file size for regular repositories.
• training_pm_nosat_150k.dat.gz
  • file with 150k traces, compressed.

TODO:

• Add example on model saving
• data exploration
  • correlations
  • energy distributions
  • position distributions (including separation distances)