Neural Networks for 3D Characterisation of AGATA Crystals

TL;DR

This paper introduces an LSTM neural network approach for 3D gamma-ray interaction localization in AGATA crystals, trained on experimental data, outperforming traditional simulation-based methods.

Contribution

The paper presents a novel LSTM-based method for 3D gamma-ray localization using experimental data and a custom loss function, improving accuracy over existing databases.

Findings

The LSTM model outperforms simulated databases in localization accuracy.

The experimental database surpasses traditional PSCS algorithm results.

A custom masked loss function enables training with incomplete data.

Abstract

Precise localisation of gamma-ray interactions is crucial for the performance of the Advanced GAmma Tracking Array (AGATA). The Pulse Shape Analysis (PSA) method used for the position estimation of gamma-ray interactions relies on a simulated signal database. The Pulse Shape Comparison Scanning (PSCS) method was used to scan AGATA crystals in order to produce an experimental database of signals. This paper presents a novel approach using Long Short-Term Memory (LSTM) neural networks to determine the 3D interaction position of gamma rays within AGATA crystals, trained on data from IPHC Strasbourg, allowing for the construction of an experimental database. A custom masked loss function is introduced to enable training with incomplete position information. The database generated by this new method outperforms the existing simulated database, and the experimental database obtained from the conventional PSCS algorithm.