Transputer self-organizing map algorithm for beam background rejection at the BELLE silicon vertex detector

A growing self-organizing map using a gravitational algorithm was implemented on a transputer to study possible separation of GEANT simulated beam background events and physics events (e+e−→B°B°, qq, ggg, 2γ,τ+τ−). A fraction of 75.0% of beam background events can be rejected, 96.9% of physics events pass the classification. The decision time is τ≤2 ms, thus the system could be used online as level 3 trigger as well as for offline data filtering purposes.