Neural Networks and other Statistical Learning Methods in High Energy and Astrophysics

Jens Zimmermann
MPI Munich

In this talk an introduction to neural networks will be given with an
emphasis on the common grounds with and the differences to other
statistical learning methods.  The underlying ideas and geometrical
interpretations of established statistical learning methods will be
discussed and compared with those of neural networks.  Clear
guidelines to the correct application and evaluation of these methods
will be given including the derivation of bias-free efficiencies and
their uncertainties.  A short introduction into the theory of
statistical learning and some very encouraging results obtained with
neural networks complete the talk.  All aspects of the talk will be
discussed with the help of examples from H1 (L2 neural network trigger
and search for instanton-induced events) and from the Cherenkov
telescope MAGIC.