Synthesis of pulses from particle detectors with a Generative Adversarial Network (GAN)

Access to the detectors and to their setup is not easy. In some cases it is required to include extra costs, long journeys to facilities, delays due to long waiting lists or even that the detectors are no longer operative. Finally, and depending on the detector, sometimes its events have a very low frequency, increasing the testing time. To avoid all this, creating realistic synthetic pulses may be a solution. With this purpose we use Generative Adversarial Networks (GANs) [1]. They are a model based on deep neural networks to create new signals that mimic the original ones. Generic GANs consist of two main neural networks: the Generator and the Discriminator. The Generator must be specialized in creating data to fool the Discriminator into accepting it. Its adversary, the Discriminator, attempts to distinguish between samples drawn from the training data and samples drawn from the Generator.