Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Generative Models for Proton Zero Degree Calorimeter Simulations in ALICE, CERN

Patryk Będkowski, Jan Dubiński, Kamil Deja, Przemysław Rokita

TL;DR

This paper addresses the computational burden of ALICE ZDC simulations by introducing a conditional Deep Convolutional GAN framework, extending SDI-GAN with an intensity regularization and an auxiliary regressor to simultaneously improve distribution fidelity and spatial localization of calorimeter showers. The method demonstrates that incorporating these regularizations yields lower Wasserstein distances on multiple calorimeter channels compared to baseline GANs, while preserving high-fidelity visuals. The contributions include the first generative simulation tailored to the Proton ZDC in ALICE, a quantitative evaluation across a large dataset, and practical improvements in speed and accuracy for detector simulations. This approach enables faster, scalable calorimeter simulations that can accelerate physics analyses and reduce reliance on expensive Monte-Carlo computations.

Abstract

Simulating detector responses is a crucial part of understanding the inner-workings of particle collisions in the Large Hadron Collider at CERN. The current reliance on statistical Monte-Carlo simulations strains CERN's computational grid, underscoring the urgency for more efficient alternatives. Addressing these challenges, recent proposals advocate for generative machine learning methods. In this study, we present an innovative deep learning simulation approach tailored for the proton Zero Degree Calorimeter in the ALICE experiment. Leveraging a Generative Adversarial Network model with Selective Diversity Increase loss, we directly simulate calorimeter responses. To enhance its capabilities in modeling a broad range of calorimeter response intensities, we expand the SDI-GAN architecture with additional regularization. Moreover, to improve the spatial fidelity of the generated data, we introduce an auxiliary regressor network. Our method offers a significant speedup when comparing to the traditional Monte-Carlo based approaches.

Deep Generative Models for Proton Zero Degree Calorimeter Simulations in ALICE, CERN

TL;DR

This paper addresses the computational burden of ALICE ZDC simulations by introducing a conditional Deep Convolutional GAN framework, extending SDI-GAN with an intensity regularization and an auxiliary regressor to simultaneously improve distribution fidelity and spatial localization of calorimeter showers. The method demonstrates that incorporating these regularizations yields lower Wasserstein distances on multiple calorimeter channels compared to baseline GANs, while preserving high-fidelity visuals. The contributions include the first generative simulation tailored to the Proton ZDC in ALICE, a quantitative evaluation across a large dataset, and practical improvements in speed and accuracy for detector simulations. This approach enables faster, scalable calorimeter simulations that can accelerate physics analyses and reduce reliance on expensive Monte-Carlo computations.

Abstract

Simulating detector responses is a crucial part of understanding the inner-workings of particle collisions in the Large Hadron Collider at CERN. The current reliance on statistical Monte-Carlo simulations strains CERN's computational grid, underscoring the urgency for more efficient alternatives. Addressing these challenges, recent proposals advocate for generative machine learning methods. In this study, we present an innovative deep learning simulation approach tailored for the proton Zero Degree Calorimeter in the ALICE experiment. Leveraging a Generative Adversarial Network model with Selective Diversity Increase loss, we directly simulate calorimeter responses. To enhance its capabilities in modeling a broad range of calorimeter response intensities, we expand the SDI-GAN architecture with additional regularization. Moreover, to improve the spatial fidelity of the generated data, we introduce an auxiliary regressor network. Our method offers a significant speedup when comparing to the traditional Monte-Carlo based approaches.
Paper Structure (9 sections, 5 equations, 3 figures, 1 table)

This paper contains 9 sections, 5 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related work
  3. Method
  4. Auxiliary regressor
  5. Diversity regularization
  6. Intensity regularization
  7. Final training objective
  8. Experiments
  9. Conclusions

Figures (3)

  • Figure 1: Architecture of the generator in used convolutional GAN across all tests.
  • Figure 2: Example of simulated responses from different methods.
  • Figure 3: Histograms of true and generated distributions of channel values. The GAN and SDI-GAN model have visible problems with underproducing high-energy responses. The implementation of additional regularization, and auxiliary regressor positively influence better alignment to true distribution, but tends to oversample the high-energy responses.