Applicability of Ann in the ARGO-YBJ experiment

We report the applicability of Artificial Neural Networks (ANN) in the ARGO-YBJ data analysis, i.e. inner or outer shower core position identification and γ-proton separation, With the MC samples from Corsika and a standard feed forward neural network, the results indicate that the rejection of outer showers induced by protons is more than 60% and the enhancement in the gamma ray sensitivity is about 37% 1 Motivation The Sino-Italian ARGO-YBJ experiment locates at YangBa-Jing (90o 31'50"E, 30o 6'38"N, 4300m a.s.l.) of Tibet, China. The main goal of the experiment is to search for Very High Energy γ point sources and HE Gamma Ray Bursts. The experimental setup is a coverage RPC carpet with an area of 97m x 103m(D'Ettorre et al., 1999) which consists of 14040 PADs. Each PAD is the detector minimum unit with 8 readout strips (i.e. the maximum number of recorded particles = 8). In order to increase the ratio of signals to noises, we have done a preliminary study on the γ-proton seperation using Artificial Neural Networks(ANN) technique(Bussino 1999). Our further Monte Carlo study indicates that the determination accuracy of event core position will affect the γ-proton identification power significantly and the key point for the determination of event core position is inner or outer event classcification. Here inner (or outer) event means the event real core located inside (or outside) of the central 71m x 74m full coverage carpet. In this note we mainly discuss on the identification power for the two classes of events.