Study of a neural approach for lower level e/g calorimeter trigger in CMS

We investigate the possibility of using Neural Network based algorithms at the lower stages of the electron/photon calorimeter trigger of CMS. The aim is to improve the background rejection obtained with currently proposed level 1 algorithms. The shower profiles are analysed in detail in the classification procedure, using the ECAL fine structure information of the trigger tower. As it can be expected, severe restrictions constrain at this level the algorithmic implementation possibilities.