The Neural Network Trigger System of the H1-Experiment at HERA
Paper: 473
Session: F (talk)
Speaker: Kiesling, Christian, Max-Planck-Institut fuer Physik, Muenchen
Keywords: ?
The Neural Network Trigger System of the H1-Experiment at HERA
Speaker:
C. Kiesling, MPI fuer Physik, Muenchen
Abstract:
We describe the concept, realisation, operation and physics
performance of a novel second-level hardware trigger for the
H1-experiment at HERA using the neural network technique. The trigger
is realized with dedicated hardware, making use of a commercially
available, massively parallel computing architecture and of
custom-designed preprocessing units. The trigger system consists of 12
independent networks, working all in parallel, each one optimized for
a specific physics reaction with high output rate from level 1, thus
requiring substantial rate reductions. The network algorithms
implemented are standard feed-forward as well as various kinds of
topological pattern recognition schemes. The trigger system is
operating successfully since the summer of 1996, providing access to
physics reactions which could not be triggered efficiently prior
to introduction of the neural network system.