The Southern Wide-field Gamma-ray Observatory (SWGO) is a next-generation ground-based gamma-ray detector array to be constructed in the Parque Astronómico near San Pedro de Atacama, Chile. It is conceived as consisting of more than 6,000 water Cherenkov detector units to cover energies ranging from 100s GeV up to the PeV scale. A major challenge in the design of SWGO is determining an optimal layout of the detector units to maximize information extraction from the recorded events. In this work, we present Southern Wide-field Gamma-ray Observatory Layout Optimizer (SWGOLO), a framework to address this challenge. SWGOLO is capable of simulating Extensive Air Shower (EAS)s, reconstructing their parameters using machine learning techniques, specifically Fully Connected Neural Network (FCNN)s, and iteratively optimizes the layout configuration with a Stochastic Gradient Descent (SGD) algorithm. The framework includes a specified objective to be maximized, which depends on the number of reconstructable/detectable EASs and inversely on the error on the reconstructed physical parameters of the EASs (energy and angular direction). Our results demonstrate the potential of SWGOLO to guide the design of an efficient and high-performing SWGO array configuration.
Southern Wide-field Gamma-ray Observatory (SWGO) è un rivelatore di raggi gamma di nuova generazione che verrà costruito nel Parque Astronómico vicino a San Pedro de Atacama, in Cile. È concepito come un insieme di oltre 6.000 unità di rivelatori Cherenkov ad acqua, sensibile a raggi gamma di energia da alcune centinaia di GeV fino alla scala dei PeV. Una delle principali sfide nella progettazione di SWGO è determinare una disposizione ottimale delle unità rivelatrici per massimizzare l’estrazione di informazione dagli eventi registrati. In questo lavoro presentiamo Southern Wide-field Gamma-ray Observatory Layout Optimizer (SWGOLO), un framework sviluppato per affrontare questa sfida. SWGOLO è in grado di simulare Extensive Air Shower (EAS), ricostruirne i parametri tramite tecniche di apprendimento automatico, in particolare Fully Connected Neural Network (FCNN), e ottimizzare iterativamente la configurazione della disposizione con un algoritmo di Stochastic Gradient Descent (SGD). Il framework include un obiettivo specifico da massimizzare, che dipende dal numero di Extensive Air Shower (EAS) rivelabili/ricostruibili e inversamente dall’errore sui parametri fisici ricostruiti degli Extensive Air Shower (EAS) (energia e direzione angolare). I nostri risultati dimostrano il potenziale di SWGOLO nel guidare la progettazione di una configurazione di SWGO efficiente e ad alte prestazioni.
Optimization of the SWGO array with gradient descent algorithms
NOYAN, ULAS FIRAT
2024/2025
Abstract
The Southern Wide-field Gamma-ray Observatory (SWGO) is a next-generation ground-based gamma-ray detector array to be constructed in the Parque Astronómico near San Pedro de Atacama, Chile. It is conceived as consisting of more than 6,000 water Cherenkov detector units to cover energies ranging from 100s GeV up to the PeV scale. A major challenge in the design of SWGO is determining an optimal layout of the detector units to maximize information extraction from the recorded events. In this work, we present Southern Wide-field Gamma-ray Observatory Layout Optimizer (SWGOLO), a framework to address this challenge. SWGOLO is capable of simulating Extensive Air Shower (EAS)s, reconstructing their parameters using machine learning techniques, specifically Fully Connected Neural Network (FCNN)s, and iteratively optimizes the layout configuration with a Stochastic Gradient Descent (SGD) algorithm. The framework includes a specified objective to be maximized, which depends on the number of reconstructable/detectable EASs and inversely on the error on the reconstructed physical parameters of the EASs (energy and angular direction). Our results demonstrate the potential of SWGOLO to guide the design of an efficient and high-performing SWGO array configuration.| File | Dimensione | Formato | |
|---|---|---|---|
|
Noyan_Ulas_Firat.pdf
accesso aperto
Dimensione
12.52 MB
Formato
Adobe PDF
|
12.52 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/92358