# Deep-Learning Jets with Uncertainties and More

## TL;DR

Bayesian neural networks enhance deep learning for jet tagging at the LHC by quantifying uncertainties related to training data, systematics, and pile-up, improving understanding and control without performance loss.

## Contribution

This paper demonstrates the application of Bayesian neural networks to jet tagging, highlighting their ability to quantify uncertainties and address systematic issues in high-energy physics.

## Key findings

- Capture statistical uncertainties from finite training samples
- Address systematics related to jet energy scale
- Improve stability against pile-up effects

## Abstract

Bayesian neural networks allow us to keep track of uncertainties, for example in top tagging, by learning a tagger output together with an error band. We illustrate the main features of Bayesian versions of established deep-learning taggers. We show how they capture statistical uncertainties from finite training samples, systematics related to the jet energy scale, and stability issues through pile-up. Altogether, Bayesian networks offer many new handles to understand and control deep learning at the LHC without introducing a visible prior effect and without compromising the network performance.