A simple guide from Machine Learning outputs to statistical criteria

TL;DR

This paper introduces methods to translate machine learning outputs into statistical significance measures, demonstrated through classification tasks using CNN, DNN, and VAE models in high-energy physics applications.

Contribution

It presents novel approaches for integrating ML training outputs with statistical significance criteria in physics studies, applicable to supervised and unsupervised learning.

Findings

Effective methods for linking ML outputs to statistical significance.

Application to high-energy physics scenarios with real data.

Demonstrated utility across different ML architectures.

Abstract

In this paper we propose ways to incorporate Machine Learning training outputs into a study of statistical significance. We describe these methods in supervised classification tasks using a CNN and a DNN output, and unsupervised learning based on a VAE. As use cases, we consider two physical situations where Machine Learning are often used: high-$p_T$ hadronic activity, and boosted Higgs in association with a massive vector boson.