Calomplification -- The Power of Generative Calorimeter Models

TL;DR

This paper demonstrates that machine-learned generative models can efficiently learn complex physics simulations, such as photon showers in calorimeters, outperforming limited training samples and reducing computational costs.

Contribution

It extends the concept of GANplification from simple Gaussian models to realistic particle physics simulations, showcasing practical benefits.

Findings

Generative models outperform limited training samples in physics simulations.

GANplification effect observed in complex calorimeter photon shower data.

Significant reduction in simulation computational costs.

Abstract

Motivated by the high computational costs of classical simulations, machine-learned generative models can be extremely useful in particle physics and elsewhere. They become especially attractive when surrogate models can efficiently learn the underlying distribution, such that a generated sample outperforms a training sample of limited size. This kind of GANplification has been observed for simple Gaussian models. We show the same effect for a physics simulation, specifically photon showers in an electromagnetic calorimeter.