Generation of Belle II Pixel Detector Background Data with a GAN

Abstract

To match the statistical precision due to the large dataset that Belle II is expected to collect, simulations that accurately describe real data are required. The effect of beam background must be considered and can be taken into account with overlaying random trigger data, but its large size is a technical challenge. This problem can be mitigated by generating beam background data with generative adversarial networks. A proof of principle is shown for the background data recorded by the pixel vertex detector.