Towards an automated data cleaning with deep learning in CRESST
Abstract
The CRESST experiment employs cryogenic calorimeters for the sensitive measurement of nuclear recoils induced by dark matter particles. The recorded signals need to undergo a careful cleaning process to avoid wrongly reconstructed recoil energies caused by pile-up and read-out artefacts. We frame this process as a time series classification task and propose to automate it with neural networks. With a data set of over one million labeled records from 68 detectors, recorded between 2013 and 2019 by CRESST, we test the capability of four commonly used neural network architectures to learn the data cleaning task. Our best performing model achieves a balanced accuracy of 0.932 on our test set. We show on an exemplary detector that about half of the wrongly predicted events are in fact wrongly labeled events, and a large share of the remaining ones have a context-dependent ground truth. We furthermore evaluate the recall and selectivity of our classifiers with simulated data. The results confirm that the trained classifiers are well suited for the data cleaning task.
Keywords
Cite
@article{arxiv.2211.00564,
title = {Towards an automated data cleaning with deep learning in CRESST},
author = {G. Angloher and S. Banik and D. Bartolot and G. Benato and A. Bento and A. Bertolini and R. Breier and C. Bucci and J. Burkhart and L. Canonica and A. D'Addabbo and S. Di Lorenzo and L. Einfalt and A. Erb and F. v. Feilitzsch and N. Ferreiro Iachellini and S. Fichtinger and D. Fuchs and A. Fuss and A. Garai and V. M. Ghete and S. Gerster and P. Gorla and P. V. Guillaumon and S. Gupta and D. Hauff and M. Ješkovský and J. Jochum and M. Kaznacheeva and A. Kinast and H. Kluck and H. Kraus and M. Lackner and A. Langenkämper and M. Mancuso and L. Marini and L. Meyer and V. Mokina and A. Nilima and M. Olmi and T. Ortmann and C. Pagliarone and L. Pattavina and F. Petricca and W. Potzel and P. Povinec and F. Pröbst and F. Pucci and F. Reindl and D. Rizvanovic and J. Rothe and K. Schäffner and J. Schieck and D. Schmiedmayer and S. Schönert and C. Schwertner and M. Stahlberg and L. Stodolsky and C. Strandhagen and R. Strauss and I. Usherov and F. Wagner and M. Willers and V. Zema and W. Waltenberger},
journal= {arXiv preprint arXiv:2211.00564},
year = {2023}
}
Comments
12 pages, 8 figures, 6 tables