CUORE Data Release for ML Applications: Pulse Shape Analysis Dataset

Abstract

We present a public dataset from the CUORE (Cryogenic Underground Observatory for Rare Events) experiment, designed to support the development and benchmarking of Artificial Intelligence and Machine Learning (AI/ML) algorithms for cryogenic calorimeter data analysis. CUORE uses TeO$_2$ cryogenic calorimeters to measure particle energy depositions as thermal fluctuations. This dataset contains thermal pulses measured during calibration data taking. Each data point is provided as a one-dimensional, time-series array corresponding to a thermal pulse, accompanied by a binary classification label to distinguish between single-pulse events and pile-up events with two or more pulses; pulse normalization parameters and relevant metadata are also included, with all data stored in HDF5 format. This data release enables the testing of supervised learning approaches to pulse shape analysis, pile-up identification, and related tasks in the context of rare-event searches with cryogenic calorimeters.