Speaker
Description
Characterizing backgrounds is an extremely important task for any dark matter search, and almost every new detector encounters unexpected backgrounds. Such was the case in LZ's first science run, in which several detector effects contributed novel backgrounds. Where the new background events exhibited abnormal PMT waveforms, these features allowed them to be identified using pulse shape information. This work focuses on the development and application of variational autoencoders (VAEs) for use on LZ's summed PMT waveforms for the purpose of identifying such anomalous backgrounds. Several VAE implementations are tested, including the use of both dense and convolutional neural networks. Findings are then compared to known detector effects which were removed from LZ's first WIMP search using pulse-shape-dependent cuts.
| Consider for long presentation | No |
|---|
