Área: DCE - Divulgação e Ensino da Física Nuclear e de Partículas Elementares
Título: SPRACE MasterClass na UFABC
Autores: André Lessa, *Eduardo Gregores, Lucio Costa, Pedro Mercadante
Instituição: UFABC
Resumo:
Isabela
Área: DCE - Divulgação e Ensino da Física Nuclear e de Partículas Elementares
Título:
Autores: *Isabela XXX, Sandra Padula
Instituição:
Resumo:
Tulio
Área: HEX - Física Experimental de Altas Energias
Título:
Autores: Eduardo Gregores, *Tulio Cardoso
Instituição: UFABC
Resumo:
Apresentações Orais
Sandra
Área: DCE - Divulgação e Ensino da Física Nuclear e de Partículas Elementares
Título:
Autores:
Instituição:
Resumo:
Thiago
Área: HEX - Física Experimental de Altas Energias
Título: Machine Learning Techniques for HL-LHC Tracking in CMS
Autores: R. Cobe, J. Fialho, R. Iope, A. Santos, S. Stanzani, T. Tomei
Instituição: SPRACE-Unesp
Resumo: The High-Luminosity LHC (HL-LHC) is the next challenge in the HEP scenario, bringing the collider’s instantaneous luminosity to 75 Hz/nb and increasing in 5 times the amount of additional pp interactions in the same or neighboring bunch crossings, referred to as pileup (PU). At an average pileup of 140, in its standard configuration, the HL-LHC will deliver to CMS a data throughput of approximately 30 GB/s, doubling to 60 GB/s at the ultimate (PU = 200) configuration. Already on 2027, the CMS experiment estimates a need of 2.2 EB of disk, 3 EB of tape and 4.4M CPI cores, with only 200 to 300/fb of data collected. At the end of the full LHC + HL-LHC experimental run, the total collected luminosity will be on the order of 3000/fb. In order to deal with the increased amount of generated data and the complexity of the simulations, new techniques and frameworks have to be deployed and/or developed. In that scenario, the Deep Neural Networks (DNN) revolution can make a significant impact on HEP. These techniques are most promising when there are both a large amount of data and a high number of features. We report on the exploration of the usage of advanced machine learning techniques for tracking at the HL-LHC, using the same dataset that was used for the TrackML Kaggle challenge.