Baikal-GVD is a large (∼ 1 km3) underwater neutrino telescope located in Lake Baikal, Russia. This work presents a neural network for reconstructing energy of muons particles born in interaction of neutrino with water. Simultaneously, the network estimates an error of it's reconstruction in terms of 1 sigma. This achieved by using special loss function, based on gaussian likelihood.
The repository contains framework for data analysis in Baikal-GVD experiment.
Repository provides code for Physical-informed neural nets:
- Linear and Convolution NN with fixed data
- Recurrent Neural Networks for work with sequential data
Preliminaries: I recommend using Linux distributions
Clone repository using git
git clone
Poetry will install proper environment for your start
# to be writtenAfter installation project can be started solely from command:
# to be written
Minimal example
# to be writtenContact me via opening issue on github or sending email matseiko.av@phystech.edu.
Use telegram for research collaboration @AlbertMac280