Abstract of Thesis presented at COPPE/UFRJ as a partial fulfillment of the requirements for the degree of Doctor of Science (D.Sc.)
Nonlinear Independent Component Analysis for Online Filtering Based on High-Energy and Highly Segmented Calorimetry
Eduardo Furtado de Simas Filho
December/2010
Advisors: |
José Manoel de Seixas
Luiz Pereira Calôba
|
Department: |
Eletrical Engineering |
ATLAS is the largest detector of the Large Hadron Collider (LHC). A large amount of information is produced in the collisions, but only a small fraction is important for characterizing interesting physics, demanding an efficient online event detection (trigger) system. Electrons are extremely important for the LHC and are immerse in a huge background noise of hadronic jets, as these last signatures present calorimeter energy deposition profile similar to electron one. Calorimeters are highly segmented energy measurement systems and in ATLAS are split into seven layers and more than 100,000 sensing elements. In this work, the nonlinear independent component analysis (NLICA) model is proposed to extract features aiming at the optimization of the ATLAS electron online neural discriminator (Neural Ringer). In order to cope with the full segmentation and granularity available, feature extraction was performed at layerlevel. Different algorithms were used for the independent components estimation. Through the proposed approach, higher discrimination efficiency was achieved, producing cleaner data for offline analysis.