Título: Denoising Autoencoder to Mitigate Crosstalk Effects in a Liquid Argon Electromagnetic Calorimeter
Autores: Ícaro Nascimento Queiroz, Antonio Carlos Lopes Fernandes Júnior, Eduardo Furtado de Simas Filho, & Marton Sandes dos Santos
Resumo: The calorimeters in ATLAS, one of the experiments carried out at CERN’s LHC, can detect the energy levels generated by the collision of various particles, thus enabling the computational reconstruction of the energy value deposited by the particle and its flight time. One of the problems observed in the sensor elements present in ATLAS is the phenomenon of crosstalk, caused by the high energy levels generated by collisions and the inclusion of a series of uncertainties into the computational reconstruction of the collision. In this article the approach suggested to mitigate these problems is the use of a Denoising Autoencoder neural network, trained on simulated data extracted from the Lorenzetti framework, developed to generate synthetic data relating to particle collisions within ATLAS. Through the use of these neural networks, we sought to carry out a neural network filtering process, using several convolutional and pooling layers, to achieve energy values closer to the true energy value.
Palavras-chave: calorimeter; lorenzetti; simulator; crosstalk; neural networks; autoencoder; denoising.
Páginas: 7
Código DOI: 10.21528/CBIC2025-1189717
Artigo em PDF: CBIC_2025_paper1189717.pdf
Arquivo BibTeX:
CBIC_2025_1189717.bib