Rejecting noise in Baikal-GVD data with neural networks

Baikal-GVD is a large ($\sim$ 1 km$^3$) underwater neutrino telescope installed in the fresh waters of Lake Baikal. The deep lake water environment is pervaded by background light, which produces detectable signals in the Baikal-GVD photosensors. We introduce a neural network for an efficient separation of these noise hits from the signal ones, stemming from the propagation of relativistic particles through the detector. The neural network has a U-net like architecture and employs temporal (causal) structure of events. On Monte-Carlo simulated data, it reaches 99% signal purity (precision) and 98% survival efficiency (recall). The benefits of using neural network for data analysis are discussed, and other possible architectures of neural networks, including graph based, are examined.