Comparing and contrastic synthetic Standard Model background samples

This is the companion repository to the paper "The Interplay of Machine Learning--based Resonant Anomaly Detection Methods", at https://arxiv.org/abs/2212.11285 (authors: Tobias Golling, Gregor Kasieczka, Claudius Krause, Radha Mastandrea, Benjamin Nachman, John Andrew Raine, Debajyoti Sengupta, David Shih, Manuel Sommerhalder).

All plots in the paper can be remade using these scripts and a dataset (to be released shortly) of synthetic Standard Model background samples from the SALAD, CATHODE, CURTAINs, and FETA methods.

For questions/comments about the code contact: [email protected]

Pipeline

Testing the synthetic samples in anomaly detection tasks

Run the script final_eval_and_scatterplot_SSS.ipynb. This will train a 5-fold binary classifier to discriminate the synthetic samples from a set ``data" (i.e. LHCO olympics background and signal). This script requires a dataset of synthetic Standard Model background samples.

Once the above script has been run, the output can be passed through the following notebooks:

• Scatterplots: use the notebook analyze_scatterplot_all_synth.ipynb.
• Calculating the overlap of shared events between methods: use the notebook analyze_scatterplot_all_synth.ipynb.
• Classifier metrics: use the notebook sample_combination_ensembling.ipynb.ipynb.

Comparing the synthetcis samples against each other

Run the script bk_comparison.ipynb. This will train a classifier to discriminate the synthetic samples against each other. Once script has been run, analyze the output with the notebook plot_bkg_comparison.ipynb.