Köln 2004 – scientific programme

Parts | Days | Selection | Search | Downloads | Help

HK: Physik der Hadronen und Kerne

HK 12: Poster Session: Instrumentation and Applications

HK 12.41: Poster

Tuesday, March 9, 2004, 13:30–15:30, Foyer

A neural network for electron/pion separation in the ALICE TRD — •Alexander Wilk — aInstitut für Kernphysik, Westfälische Wilhelms-Universität, Münster, Germany

One of the main features of the ALICE Transition Radiation Detector (TRD) is the identification of electrons with a momentum p > 1 GeV/c. In this poster we present an alternative method for the separation of electrons and pions in the ALICE TRD.

The “classical” methods of e/π separation employing a likelihood on integrated energy deposit and a bidimensional likelihood on energy deposit and position of the largest cluster achieve at an electron efficiency of 90% a pion rejection factor better than 100 for six TRD layers[1]. These methods use only a limited part of the information measured with ALICE TRD. The amplitude measurement of each time bin can be further exploited provided the correlations are properly taken into account. Here we show the performance of a neural network with regard to e/π separation which was trained using testbeam data. We use SNNS[2] to test different setups, and compare it to the “classical” methods.

Supported by BMBF and GSI

[1] A.Andronic et al.(for the ALICE Collaboration), GSI Scientific Report 2002, Prototype tests for the ALICE TRD

[2] SNNS-Stuttgart Neural Network Simulator,http://www-ra.informatik.uni-tuebingen.de/SNNS/