The detection of Gamma-Ray Bursts (GRBs) using spaceborne X/gamma-ray photon detectors depends on a reliable background count rate estimate. This study focuses on evaluating a data-driven background estimator based on a neural network designed to adapt to various X/gamma-ray space telescopes. Three trials were conducted to assess the effectiveness and limitations of the proposed estimator. Firstly, quantile regression was employed to obtain an estimation with a confidence range prediction. Secondly, we assessed the performance of the neural network, emphasizing that a dataset of four months is sufficient for training. We tested its adaptability across various temporal contexts, identified its limitations and recommended re-training for each specific period. Thirdly, utilizing Explainable Artificial Intelligence (XAI) techniques, we delved into the neural network output, determining distinctions between a network trained during solar maxima and one trained during solar minima. This entails conducting a thorough analysis of the neural network behavior under varying solar conditions.
Enhancing Gamma-Ray Burst Detection: Evaluation of Neural Network Background Estimator and Explainable AI Insights
Crupi R.
;Dilillo G.;Della Casa G.;Vacchi A.
2024-01-01
Abstract
The detection of Gamma-Ray Bursts (GRBs) using spaceborne X/gamma-ray photon detectors depends on a reliable background count rate estimate. This study focuses on evaluating a data-driven background estimator based on a neural network designed to adapt to various X/gamma-ray space telescopes. Three trials were conducted to assess the effectiveness and limitations of the proposed estimator. Firstly, quantile regression was employed to obtain an estimation with a confidence range prediction. Secondly, we assessed the performance of the neural network, emphasizing that a dataset of four months is sufficient for training. We tested its adaptability across various temporal contexts, identified its limitations and recommended re-training for each specific period. Thirdly, utilizing Explainable Artificial Intelligence (XAI) techniques, we delved into the neural network output, determining distinctions between a network trained during solar maxima and one trained during solar minima. This entails conducting a thorough analysis of the neural network behavior under varying solar conditions.File | Dimensione | Formato | |
---|---|---|---|
galaxies-12-00012-v2.pdf
accesso aperto
Tipologia:
Versione Editoriale (PDF)
Licenza:
Creative commons
Dimensione
1.48 MB
Formato
Adobe PDF
|
1.48 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.