Nell'ambito della costruzione di ASTRI Mini-Array, AMAMI si propone di potenziare il lavoro in corso del gruppo OA Roma per lo sviluppo della pipeline Data Reduction, con una pipeline interamente ridisegnata basata su metodi di deep learning. L'analisi standard si basa su tecniche come Random Forests e Ensemble Methods, utilizzando i parametri dell'immagine come input. Le prestazioni di ricostruzione possono probabilmente essere migliorate con modelli di deep learning che utilizzano immagini come input.
Inoltre, AMAMI esplorerà anche l'intrigante possibilità di utilizzare i recentissimi modelli diffusivi sia per generare eventi sintetici, per aumentare l'efficienza dell'attuale catena di simulazione Monte Carlo, sia per rimuovere il rumore da immagini calibrate.
In the context of the building of the ASTRI Mini-Array, an array of nine 4-m class IACTs in Tenerife (Spain), AMAMI proposes to boost the current work of the OA Roma group for the development of the Data Reduction pipeline, with an entirely redesigned pipeline based on deep learning methods. The standard analysis is based on techniques such as Random Forests and Ensemble Methods, using image parameters as inputs. Reconstruction performance can likely be improved with deep learning models using images coupled with temporal information as input.
Moreover, AMAMI will also explore the intriguing chance to use the very recent diffusive models both to generate synthetic events, to increase the efficiency of the current Monte Carlo simulation chain, and to denoise calibrated images.
Fenomeni non termici, raggi cosmici e astroparticelle
Tecnologie Informatiche e software
Tecnologie per osservazioni da Terra
Infrastrutture da Terra (utilizzo)
Team Summary
15. Personale INAF coinvolto
Numero di partecipanti INAF al progetto: 5
Struttura
Nfte
N0
TI 2023
TI 2024
TI 2025
TD 2023
TD 2024
TD 2025
Nex
Extra
O.A. ROMA
1
0
0.10
0.10
0.10
0.00
0.00
0.00
0
0.00
IAPS ROMA
0
0
0.00
0.00
0.00
0
0
0
0
0.00
Totali
1
0
0.10
0.10
0.10
0.00
0.00
0.00
0
0.00
Fondi a sostegno
21. Totale fondi a disposizione (dato aggregato, k€)
Certi 2023
Certi 2024
Certi 2025
Presunti 2023
Presunti 2024
Presunti 2025
0
0
0
0
0
0
Produzione scientifica e tecnologica
22. Produzione scientifica e tecnologica - Highlights
#
DOI
Descrizione
Azione
1
https://doi.org/10.48550/arXiv.2212.10281
Cats vs Dogs, Photons vs Hadrons, Francesco Visconti
In gamma ray astronomy with Cherenkov telescopes, machine learning models are needed to guess what kind of particles generated the detected light, and their energies and directions. The focus in this work is on the classification task, training a simple convolutional neural network suitable for binary classification (as it could be a cats vs dogs classification problem), using as input uncleaned images generated by Montecarlo data for a single ASTRI telescope. Results show an enhanced discriminant power with respect to classical random forest methods.
2
https://doi.org/10.48550/arXiv.2209.00796
Diffusion Models: A Comprehensive Survey of Methods and Applications
Ling Yang, Zhilong Zhang, Yang Song, Shenda Hong, Runsheng Xu, Yue Zhao, Wentao Zhang, Bin Cui, Ming-Hsuan Yang
Diffusion models have emerged as a powerful new family of deep generative models with record-breaking performance in many applications, including image synthesis, video generation, and molecule design. In this survey, we provide an overview of the rapidly expanding body of work on diffusion models, categorizing the research into three key areas: efficient sampling, improved likelihood estimation, and handling data with special structures. We also discuss the potential for combining diffusion models with other generative models for enhanced results. We further review the wide-ranging applications of diffusion models in fields spanning from computer vision, natural language generation, temporal data modeling, to interdisciplinary applications in other scientific disciplines. This survey aims to provide a contextualized, in-depth look at the state of diffusion models, identifying the key areas of focus and pointing to potential areas for further exploration
3
https://doi.org/10.48550/arXiv.2211.12444
Can denoising diffusion probabilistic models generate realistic astrophysical fields?
Nayantara Mudur, Douglas P. Finkbeiner
Score-based generative models have emerged as alternatives to generative adversarial networks (GANs) and normalizing flows for tasks involving learning and sampling from complex image distributions. In this work we investigate the ability of these models to generate fields in two astrophysical contexts: dark matter mass density fields from cosmological simulations and images of interstellar dust. We examine the fidelity of the sampled cosmological fields relative to the true fields using three different metrics, and identify potential issues to address. We demonstrate a proof-of-concept application of the model trained on dust in denoising dust images. To our knowledge, this is the first application of this class of models to the interstellar medium.
4
https://doi.org/10.1103/PhysRevLett.120.042003
Accelerating Science with Generative Adversarial Networks: An Application to 3D Particle Showers in Multilayer Calorimeters
Michela Paganini, Luke de Oliveira, and Benjamin Nachman
Physicists at the Large Hadron Collider (LHC) rely on detailed simulations of particle collisions to build expectations of what experimental data may look like under different theoretical modeling assumptions. Petabytes of simulated data are needed to develop analysis techniques, though they are expensive to generate using existing algorithms and computing resources. The modeling of detectors and the precise description of particle cascades as they interact with the material in the calorimeter are the most computationally demanding steps in the simulation pipeline. We therefore introduce a deep neural network-based generative model to enable high-fidelity, fast, electromagnetic calorimeter simulation. There are still challenges for achieving precision across the entire phase space, but our current solution can reproduce a variety of particle shower properties while achieving speedup factors of up to 100000×. This opens the door to a new era of fast simulation that could save significant computing time and disk space, while extending the reach of physics searches and precision measurements at the LHC and beyond.
5
https://doi.org/10.1109/CBMI50038.2021.9461918
First Full-Event Reconstruction from Imaging Atmospheric Cherenkov Telescope Real Data with Deep Learning
Mikaël Jacquemont; Thomas Vuillaume; Alexandre Benoit; Gilles Maurin; Patrick Lambert; Giovanni Lamanna
The Cherenkov Telescope Array is the future of ground-based gamma-ray astronomy. Its first prototype telescope built on-site, the Large Size Telescope 1, is currently under commissioning and taking its first scientific data. In this paper, we present for the first time the development of a full-event reconstruction based on deep convolutional neural networks and its application to real data. We show that it outperforms the standard analysis, both on simulated and on real data, thus validating the deep approach for the CTA data analysis. This work also illustrates the difficulty of moving from simulated data to actual data.
6
https://doi.org/10.22323/1.395.0730
Titolo: Reconstruction of stereoscopic CTA events using deep learning with CTLearn
Autori:Tjark Miener and Daniel Nieto and Aryeh Brill and Samuel Timothy Spencer and Jose Luis Contreras
Publisher:Sissa Medialab
Anno pubblicazione:2021
Abstract: The Cherenkov Telescope Array (CTA), conceived as an array of tens of imaging atmospheric Cherenkov telescopes (IACTs), is an international project for a next-generation ground-based gamma-ray observatory, aiming to improve on the sensitivity of current-generation instruments a factor of five to ten and provide energy coverage from 20 GeV to more than 300 TeV. Arrays of IACTs probe the very-high-energy gamma-ray sky. Their working principle consists of the simultaneous observation of air showers initiated by the interaction of very-high-energy gamma rays and cosmic rays with the atmosphere. Cherenkov photons induced by a given shower are focused onto the camera plane of the telescopes in the array, producing a multi-stereoscopic record of the event. This image contains the longitudinal development of the air shower, together with its spatial, temporal, and calorimetric information. The properties of the originating very-high-energy particle (type, energy, and incoming direction) can be inferred from those images by reconstructing the full event using machine learning techniques. In this contribution, we present a purely deep-learning driven, full-event reconstruction of simulated, stereoscopic IACT events using CTLearn. CTLearn is a package that includes modules for loading and manipulating IACT data and for running deep learning models, using pixel-wise camera data as input.
Titolo: Application of deep learning methods to analysis of imaging atmospheric Cherenkov telescopes data
Autori:I. Shilon and M. Kraus and M. Büchele and K. Egberts and T. Fischer and T.L. Holch and T. Lohse an ....
Publisher:Elsevier BV
Rivista: Astroparticle Physics
Anno pubblicazione:2019
Abstract: Ground based γ-ray observations with Imaging Atmospheric Cherenkov Telescopes (IACTs) play a significant role in the discovery of very high energy (E > 100 GeV) γ-ray emitters. The analysis of IACT data demands a highly efficient background rejection technique, as well as methods to accurately determine the position of its source in the sky and the energy of the recorded γ-ray. We present results for background rejection and signal direction reconstruction from first studies of a novel data analysis scheme for IACT measurements. The new analysis is based on a set of Convolutional Neural Networks (CNNs) applied to images from the four H.E.S.S. phase-I telescopes. As the H.E.S.S. cameras pixels are arranged in a hexagonal array, we demonstrate two ways to use such image data to train CNNs: by resampling the images to a square grid and by applying modified convolution kernels that conserve the hexagonal grid properties.
The networks were trained on sets of Monte-Carlo simulated events and tested on both simulations and measured data from the H.E.S.S. array. A comparison between the CNN analysis to current state-of-the-art algorithms reveals a clear improvement in background rejection performance. When applied to H.E.S.S. observation data, the CNN direction reconstruction performs at a similar level as traditional methods. These results serve as a proof-of-concept for the application of CNNs to the analysis of events recorded by IACTs.
8
https://doi.org/10.48550/arXiv.2211.16009
Titolo: The performance of the MAGIC telescopes using deep convolutional neural networks with CTLearn
Autori: T. Miener, D. Nieto, R. López-Coto, J. L. Contreras, J. G. Green, D. Green, E. Mariotti
Abstract: The Major Atmospheric Gamma Imaging Cherenkov (MAGIC) telescope system is located on the Canary Island of La Palma and inspects the very high-energy (VHE, few tens of GeV and above) gamma-ray sky. MAGIC consists of two imaging atmospheric Cherenkov telescopes (IACTs), which capture images of the air showers originating from the absorption of gamma rays and cosmic rays by the atmosphere, through the detection of Cherenkov photons emitted in the shower. The sensitivity of IACTs to gamma-ray sources is mainly determined by the ability to reconstruct the properties (type, energy, and arrival direction) of the primary particle generating the air shower. The state-of-the-art IACT pipeline for shower reconstruction is based on the parameterization of the shower images by extracting geometric and stereoscopic features and machine learning algorithms like random forest or boosted decision trees. In this contribution, we explore deep convolutional neural networks applied directly to the pixelized images of the camera as a promising method for IACT full-event reconstruction and present the performance of the method on observational data using CTLearn, a package for IACT event reconstruction that exploits deep learning.
9
https://doi.org/10.48550/arXiv.2006.14927
Titolo: Optimizing Cherenkov photons generation and propagation in CORSIKA for CTA Monte-Carlo simulations
Autori: Luisa Arrabito, Konrad Bernlöhr, Johan Bregeon, Matthieu Carrère, Adnane Khattabi, Philippe Langlois, David Parello, Guillaume Revy
Abstract: COsmic Ray SImulations for KAscade) is a program for detailed simulation of extensive air showers initiated by high energy cosmic ray particles in the atmosphere, and is used today by almost all the major instruments that aim at measuring primary and secondary cosmic rays on the ground. The Cherenkov Telescope Array (CTA), currently under construction, is the next-generation instrument in the field of very-high-energy gamma-ray astronomy. Detailed CORSIKA Monte Carlo simulations will be regularly performed in parallel to CTA operations to estimate the instrument response functions, necessary to extract the physical properties of the cosmic sources from the measurements during data analysis. The estimated CPU time associated with these simulations is very high, of the order of 200 million HS06 hours per year. Code optimization becomes a necessity towards fast productions and limited costs. We propose in this paper multiple code transformations that aim to facilitate automatic vectorization done by the compiler, ensuring minimal external libraries requirement and high hardware portability.
10
https://doi.org/10.1117/12.2629362
Titolo: The data processing, simulation, and archive systems of the ASTRI Mini-Array project
Autori:Saverio Lombardi and Fabrizio Lucarelli and Ciro Bigongiari and Stefano Gallozzi and Martina Cardil ....
Publisher:SPIE
Anno pubblicazione:2022
The ASTRI Mini-Array is an international project led by the Italian National Institute for Astrophysics (INAF) to build and operate an array of nine 4-m class Imaging Atmospheric Cherenkov Telescopes (IACTs) at the Observatorio del Teide (Tenerife, Spain). The system is designed to perform deep observations of the galactic and extragalactic gamma-ray sky in the TeV and multi-TeV energy band, with important synergies with other ground-based gamma-ray facilities in the Northern Hemisphere and space-borne telescopes. As part of the overall software system, the ASTRI (Astrofisica con Specchi a Tecnologia Replicante Italiana) Team is developing dedicated systems for Data Processing, Simulation, and Archive to achieve effective handling, dissemination, and scientific exploitation of the ASTRI Mini-Array data. Thanks to the high-speed network connection available between Canary Islands and Italy, data acquired on-site will be delivered to the ASTRI Data Center in Rome immediately after acquisition. The raw data will be then reduced and analyzed by the Data Processing System up to the generation of the final scientific products. Detailed Monte Carlo simulated data will be produced by the Simulation System and exploited in several data processing steps in order to achieve precise reconstruction of the physical characteristics of the detected gamma rays and to reject the overwhelming background due to charged cosmic rays. The data access at different user levels and for different use cases, each one with a customized data organization, will be provided by the Archive System. In this contribution we present these three ASTRI Mini-Array software systems, focusing on their main functionalities, components, and interfaces.
<p><span id="docs-internal-guid-c43bbb8c-7fff-830b-f807-ac702d2a73d5"><span style="font-size: 11pt; font-family: Consolas, sans-serif; color: rgb(0, 0, 0); background-color: transparent; font-variant-numeric: normal; font-variant-east-asian: normal; font-variant-alternates: normal; vertical-align: baseline; white-space: pre-wrap;">INAF, an active member of the MAGIC\cite{MAGIC} collaboration since 2008, is currently leading the ASTRI project\cite{ASTRI} and is a founding member of the CTA consortium\cite{CTA}. These partnerships have enabled INAF to acquire a relevant role in the field of gamma-ray astronomy and have allowed the INAF researchers to become experts in data-analysis and data-simulation for gamma-ray telescopes as well as in the development of data reduction and analysis pipelines. All members of AMAMI are experienced in different aspects of the afore-mentioned activities and complement each other to form a team perfectly suited to the project goal. INAF's involvement in this field of research is expected to grow even further in the coming years due to its leading role in the PNRR-CTA+ project. The huge effort carried on by INAF in the development and construction of new generation IACT arrays, ASTRI and CTA, could be poorly rewarded if INAF members would not consolidate their expertise in the IACT data analysis mastering innovative and promising techniques based on deep learning methods that many research groups are approaching.</span></span><br></p>
<p>CTA; ASTRI; MAGIC; HESS; VERITAS</p>
15. Team members, Informazioni generali
15. Personale INAF coinvolto
#
Nome
E-mail
Struttura
TI
Qualifica
Ruolo nel Progetto
FTE Impegnate (2023/2024/2025)
FTE Presunte (2023/2024/2025)
Extra
1
saverio.lombardi
saverio.lombardi@inaf.it
O.A. ROMA
Y
RICERCATORE
Scientist
X X X
X X X
X
2
michele.mastropietro
michele.mastropietro@inaf.it
O.A. ROMA
N
ASSEGNISTA
Scientist - Engineer
X X X
X X X
X
3
ciro.bigongiari
ciro.bigongiari@inaf.it
O.A. ROMA
Y
RICERCATORE
Scientist
X X X
X X X
X
4
martina.cardillo
martina.cardillo@inaf.it
IAPS ROMA
Y
RICERCATORE
Scientist
X X X
X X X
X
5
francesco.visconti
francesco.visconti@inaf.it
O.A. ROMA
Y
TECNOLOGO
PI - Software Architect and Engineer
X X X
X X X
X
16. Personale Associato INAF coinvolto
#
Nome
E-mail
Struttura
TI
Qualifica
Ruolo nel Progetto
FTE Impegnate (2023/2024/2025)
FTE Presunte (2023/2024/2025)
Extra
21. Fondi a Sostegno Iniziativa
Richiesta di finanziamento sottomessa nell'ambito della call 2023 per i Techno Grant INAF
