BEGIN:VCALENDAR VERSION:2.0 PRODID:-//CERN//INDICO//EN BEGIN:VEVENT SUMMARY:Welcome DTSTART:20250128T140000Z DTEND:20250128T143000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10703@events.icecube.wisc.edu DESCRIPTION:Speakers: Frank Schroeder (University of Delaware / Karlsruhe Institute of Technology)\n\nhttps://events.icecube.wisc.edu/event/243/cont ributions/10703/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10703/ END:VEVENT BEGIN:VEVENT SUMMARY:The Radar Echo Telescope for Cosmic rays DTSTART:20250129T201000Z DTEND:20250129T203000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10698@events.icecube.wisc.edu DESCRIPTION:Speakers: Krishna Nivedita Gopinath (Radboud University)\, Kat ie Mulrey (University of Delaware)\n\nThe Radar Echo Telescope for Cosmic Rays (RET-CR) was deployed this year at the high-altitude Summit Station i n Greenland. Its primary goal is to detect in-ice continuations of high-en ergy cosmic-ray-induced air showers using the radar echo method. Successfu lly detecting in-ice cosmic-ray signals through this technique would provi de significant insights and serve as a foundation for the establishment of the Radar Echo Telescope for Neutrinos (RET-N).\nThis talk will focus on the radar echo technique\, analysis of RET-CR surface station data for rec onstruction of key parameters\, including primary energy\, arrival directi on\, and core positions. It would also involve studying the combined askar yan and radar signals.\n\nhttps://events.icecube.wisc.edu/event/243/contri butions/10698/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10698/ END:VEVENT BEGIN:VEVENT SUMMARY:Teaching AI and ML to phsyics students DTSTART:20250128T143000Z DTEND:20250128T153000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10758@events.icecube.wisc.edu DESCRIPTION:Speakers: Federica Bianco (University of Delaware)\n\nArtifici al Intelligence (AI) is pervading all aspects of our society\, and can be leveraged to support and facilitate scientific discovery. However\, teachi ng AI in the rapidly evolving AI landscape is difficult. I will cover the essential tools and concepts that physics students should familiarize with in their course work to be on track to master AI applications in physics and astronomy\, and methods for teaching them\, including leveraging AI wh en writing code.\n\nhttps://events.icecube.wisc.edu/event/243/contribution s/10758/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10758/ END:VEVENT BEGIN:VEVENT SUMMARY:Application of graph networks to a next generation wide-field gamm a-ray observatory in the southern sky DTSTART:20250130T170000Z DTEND:20250130T173000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10665@events.icecube.wisc.edu DESCRIPTION:Speakers: Martin Schneider (Erlangen Centre for Astroparticle Physics (ECAP)\, Friedrich-Alexander-Universität Erlangen-Nürnberg )\, F ranziska Leitl (Erlangen Centre for Astroparticle Physics (ECAP)\, Friedri ch-Alexander-Universität Erlangen-Nürnberg)\, Christopher van Eldik (Erl angen Centre for Astroparticle Physics (ECAP)\, Friedrich-Alexander-Univer sität Erlangen-Nürnberg )\, Stefan Funk (Erlangen Centre for Astropartic le Physics (ECAP)\, Friedrich-Alexander-Universität Erlangen-Nürnberg )\ , Jonas Glombitza (Erlangen Centre for Astroparticle Physics (ECAP)\, Frie drich-Alexander-Universität Erlangen-Nürnberg )\n\nWater-Cherenkov detec tors have long proven their importance for the research of high energetic gamma rays in numerous experiments in the Northern Hemisphere.\nThe Southe rn Wide-field Gamma-ray Observatory (SWGO) will be the first observatory u sing this technology in the Southern Hemisphere to observe gamma-ray emiss ion in an energy range of 100s of GeV up to the PeV scale.\nThe proposed l ayout will enable observations of the galactic center with a wide field of view and a very high-duty cycle\, complementing the Cherenkov Telescope A rray (CTA).\nThe challenge of precision observation lies in rejecting cosm ic-ray backgrounds and accurately reconstructing primary energy\, using on ly the air shower footprint captured by the detector.\nWith new machine-le arning techniques advancing in recent times\, we propose a novel approach based on graph neural networks (GNNs) to improve background rejection and energy reconstruction for a next-generation observatory.\nWe selected a GN N-based approach to leverage the capabilities of convolutional neural netw orks while offering the flexibility needed for event reconstruction across the extensive energy range of SWGO.\n\nIn this talk\, we introduce the de sign of the proposed GNN\, describe the details of the application to a te st configuration of SWGO\, and present the obtained results for $\\gamma\\ \,/ \\\,$hadron separation and energy reconstruction.\nComparing our resul ts to current state-of-the-art approaches\, we find that our proposed algo rithm outperforms hand-designed classification algorithms and observables in background suppression\, and improves the energy resolution compared to state-of-the-art methods.\n\nhttps://events.icecube.wisc.edu/event/243/co ntributions/10665/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10665/ END:VEVENT BEGIN:VEVENT SUMMARY:Towards improving efficiency of machine learning techniques in neu trino telescopes DTSTART:20250128T210000Z DTEND:20250128T214500Z DTSTAMP:20260521T143700Z UID:indico-contribution-10666@events.icecube.wisc.edu DESCRIPTION:Speakers: Felix Yu\n\nNeutrino telescopes detect rare interact ions of particles produced from some of the most extreme environments in t he Universe. Given the rate of backgrounds\, these telescopes amass an eno rmous quantity of large variance\, high-dimensional data. These attributes create substantial challenges for analyzing and reconstructing neutrinos\ , particularly when utilizing machine learning (ML) techniques. In this ta lk\, I will present methods to efficiently manage and process these events using ML techniques\, while preserving as much information as possible. T hese methods include employing autoencoder networks to generate compact re presentations of high-dimensional data associated to neutrino events and u tilizing sparse networks to substantially reduce memory usage and runtime. The ultimate aim of these efforts is to enable high-quality ML-based reco nstructions during the earlier stages of data processing.\n\nhttps://event s.icecube.wisc.edu/event/243/contributions/10666/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10666/ END:VEVENT BEGIN:VEVENT SUMMARY:Denoising Radio Pulses from Air Showers Using Machine Learning Met hods (Remote) DTSTART:20250129T170000Z DTEND:20250129T171500Z DTSTAMP:20260521T143700Z UID:indico-contribution-10681@events.icecube.wisc.edu DESCRIPTION:Speakers: Zhisen Lai (SFSU)\, Aurélien Benoit-Lévy (CEA-List )\, Arsene Ferriere (CNRS)\, Oscar Macias (San Francisco State University) \, Claire Guepin\n\nThe Giant Radio Array for Neutrino Detection (GRAND) a ims to detect radio signals from extensive air showers caused by ultra-hig h-energy cosmic particles. Galactic\, instrumental\, and anthropogenic no ise are expected to contaminate these signals.\nTo address this problem\, we propose training an unsupervised convolutional network known as an auto encoder. This network is used to learn a coded representation of the data and remove specific features from it. This denoiser is trained using reali stic air-shower simulations generated by CoREAS and ZHAireS\, which are sp ecifically designed to closely resemble the signals detected by GRAND. In this talk\, we will present details about our machine-learning model and p reliminary results on the sensitivity gain obtained when our denoising alg orithm is applied to realistically simulated noisy GRAND signals of varyin g signal-to-noise ratios.\n\nhttps://events.icecube.wisc.edu/event/243/con tributions/10681/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10681/ END:VEVENT BEGIN:VEVENT SUMMARY:Optimizing a Cosmic-ray Energy Estimator with Machine learning for the HAWC observatory (Remote) DTSTART:20250130T200000Z DTEND:20250130T202000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10667@events.icecube.wisc.edu DESCRIPTION:Speakers: Ibrahim Torres (INAOE)\, Tomás Capistrán (Universi tà degli Studi di Torino & INFN Sezione di Torino)\, Jorge Jaimes-Teherá n (Universidad Industrial de Santander)\, for the HAWC Collaboration\n\nS ituated at an elevation of 4\,100 meters a.s.l. in Puebla\, Mexico\, the H igh-Altitude Water Cherenkov (HAWC) gamma-ray observatory detects TeV gamm a-rays from astrophysical sources. Additionally\, it gathers substantial d ata on hadronic air showers\, expanding HAWC’s research capabilities to explore cosmic rays with energies from 1 TeV to 1 PeV. The initial energy estimation method optimized for cosmic rays enabled the analysis of the a nisotropy and composition of the cosmic rays. However\, recent improvement s in HAWC reconstruction algorithms have pointed out the need for an impr oved energy estimator of hadronic EAS. To this end\, it is important to e xplore more sophisticated methods for cosmic-ray energy reconstruction. In this work\, we present preliminary results of the implementation of machi ne learning techniques for predicting the energy of cosmic-ray-induced eve nts in HAWC.\n\nhttps://events.icecube.wisc.edu/event/243/contributions/10 667/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10667/ END:VEVENT BEGIN:VEVENT SUMMARY:Graph Neural Networks for Photon Search with the Underground Muon Detector of the Pierre Auger Observatory (Remote) DTSTART:20250130T190000Z DTEND:20250130T192000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10687@events.icecube.wisc.edu DESCRIPTION:Speakers: Ezequiel Rodriguez (ITeDA-KIT)\n\nUltra-high-energy (UHE) photons are expected as by-products of cosmic-ray acceleration\, pro pagation\, or decay of super-heavy dark matter particles. Predicted diffus e photon fluxes are usually several orders of magnitude below the UHE cosm ic-ray flux. This contribution presents a method for discriminating photon -initiated air showers in the overwhelming cosmic-ray background with the Pierre Auger Observatory. The method leverages information from both the S urface Detector (SD)\, consisting of water-Cherenkov detectors (WCDs) and the Underground Muon Detector (UMD). We use graph neural networks\, that a llow the encoding of the input information acquired by the SD and UMD. The approach is particularly suitable for handling the irregular geometries o f the SD and UMD arrays\, where stations may be temporarily missing due to technical issues. Using simulations\, the performance estimates indicate that the method has a strong potential for identifying photons at UHE.\n\n https://events.icecube.wisc.edu/event/243/contributions/10687/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10687/ END:VEVENT BEGIN:VEVENT SUMMARY:Deep Learning applied to CTAO LST-1 and the difficulty to go from simulated to real data (Remote) DTSTART:20250128T171500Z DTEND:20250128T173000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10689@events.icecube.wisc.edu DESCRIPTION:Speakers: Thomas Vuillaume (LAPP\, Univ. Savoie Mont-Blanc\, C NRS)\, Michaël Dell'aiera (LAPP\, Univ. Savoie Mont-Blanc\, CNRS)\, Alexa ndre Benoit (LISTIC\, Univ. Savoie Mont-Blanc)\n\nGammaLearn is a project developing deep learning solutions for the Cherenkov Telescope Array Obser vatory (CTAO) data analysis. Its first application is event reconstruction based on images acquired by the Large-Sized Telescope (LST-1)\, currently under commissioning at La Palma.\nIn this talk\, we present a review of t he project: the architecture $\\gamma$-PhysNet we have developed to tackle this multi-task problem\, the results obtained on simulated and real data \, as well as solutions developed to compensate for some of the issues ari sing from data vs simulation discrepancies.\n\nhttps://events.icecube.wisc .edu/event/243/contributions/10689/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10689/ END:VEVENT BEGIN:VEVENT SUMMARY:Mass composition study with machine learning on KASCADE archival d ata (Remote) DTSTART:20250130T140000Z DTEND:20250130T142000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10669@events.icecube.wisc.edu DESCRIPTION:Speakers: Nikita Petrov (Novosibirsk State University)\, Vladi mir Sotnikov\, Mikhail Kuznetsov\, Ivan Plokhikh\n\nDespite many experimen ts in the 1–100 PeV range\, accurately measuring mass component spectra in this energy region remains challenging. Discrepancies between experimen ts are attributed to factors including the choice of the hadronic interact ion model used.\nIn this study\, we present a reanalysis of archival data from the KASCADE experiment\, which recorded extensive air showers from 19 96 to 2013. This analysis uses a novel approach to measure the energy spec tra of five mass components (protons\, helium\, carbon\, silicon\, and iro n)\, based on event-by-event mass-type reconstruction using a convolutiona l neural network. \nThe systematic uncertainties\, which were lower than i n the last original KASCADE study\, as well as the corresponding uncertain ties of the IceTop and TALE experiments\, were also investigated. Furtherm ore\, the uncertainties associated with the use of different post-LHC hadr onic interaction models (QGSJet-II.04\, EPOS-LHC\, Sibyll 2.3c) were exami ned.\nOur findings show a marked excess of the proton component flux compa red to the latest original KASCADE results. We demonstrate\, with the high est statistical significance\, a knee in the proton (~4.4 PeV) and helium (~11 PeV) components. Additionally\, we observe a hint of hardening (~4.5 PeV) in the iron spectrum\, which can be interpreted as analogous to the p roton hardening (~166 TeV) observed in the GRAPES-3 experiment.\n\nhttps:/ /events.icecube.wisc.edu/event/243/contributions/10669/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10669/ END:VEVENT BEGIN:VEVENT SUMMARY:Generative Neural Networks for Simulating Radio Emission from Air Showers (Remote) DTSTART:20250129T160000Z DTEND:20250129T162000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10674@events.icecube.wisc.edu DESCRIPTION:Speakers: Pranav Sampathkumar (IAP\, KIT)\, Tim Huege (Karlsru he Institute of Technology)\n\nThe simulations of radio emission from EAS are essential for reconstructing various shower parameters from the measur ed radio signals. As bigger experiments use more and more antennas\, the c omputational cost of these simulations gets prohibitively large. These sim ulations also scale exponentially with higher primary energies and linearl y with the number of antennas. Thus there is a need to interpolate and gen erate radio signals across various energy ranges and antenna positions.\n\ nIn this work\, we present a novel neural network which can predict radio pulses for the AERA setup using several shower parameters and antenna posi tions as input. The results which showcase the pulses generated by the net work compared to the CoREAS simulations\, are presented. The network’s a bility to also get the fluence pattern and the total radiation energy is s hown along with its performance benchmarks. Finally\, the network is used in a simplistic Xmax reconstruction procedure to show the viability of the se generated pulses for Xmax reconstruction.\n\nhttps://events.icecube.wis c.edu/event/243/contributions/10674/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10674/ END:VEVENT BEGIN:VEVENT SUMMARY:Gamma/Hadron Separation using Machine Learning Methods with the Ic eAct Telescopes DTSTART:20250129T213000Z DTEND:20250129T215000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10671@events.icecube.wisc.edu DESCRIPTION:Speakers: Logan Molchany (University of Delaware)\, Matthias P lum (South Dakota School of Mines and Technology)\n\nThe IceCube Neutrino Observatory\, located at the South Pole\, is a multi-component detector ar ray capable of observing cosmic-rays on the TeV to EeV scale. In addition to the InIce component\, and the surface component IceTop\, three new Imag ing Air Cherenkov Telescopes\, called IceAct\, were installed. One of the primary goals of the IceAct telescopes is to search for high-energy photon s in the Southern Sky. To do so\, Gamma/Hadron separation is done by usin g modern machine learning methods alongside a hybrid Hillas analysis which uses both the Hillas parameters alongside InIce parameters. This approac h geometrically parameterizes the ellipse formed by the images on the IceA ct cameras alongside the total charge deposition at various layers in ice\ , as well as using reconstructed muon bundle energy loss as model features . Various classification and regression models are used to reconstruct th e energy and type of the primary cosmic-ray\, as well as a final meta-mode ling approach that aggregates the predictions from all used models in a so -called "stacking method".\n\n\nThis contribution will provide a prelimina ry look into the sensitivity of the current machine learning model/ stacki ng method\, used for distinguishing photons from the cosmic ray background .\n\nhttps://events.icecube.wisc.edu/event/243/contributions/10671/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10671/ END:VEVENT BEGIN:VEVENT SUMMARY:Reconstructing the Direction of Ultra-High-Energy Cosmic Rays Usin g a Simulation-Based Inference Method DTSTART:20250129T195500Z DTEND:20250129T201000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10682@events.icecube.wisc.edu DESCRIPTION:Speakers: Matthew Ho\, Aurélien Benoit-Lévy (CEA-List)\, Os car Macias (San Francisco State University)\, Zach Mason (SFSU)\n\nGRAND ( Giant Radio Array for Neutrino Detection) is a proposed next-generation ob servatory designed to detect ultra-high-energy (UHE) cosmic particles. It aims to accomplish this by identifying the radio signals generated when th ese particles interact with the atmosphere and Earth's magnetic field. We present a novel pipeline utilizing simulation-based inference (SBI) method s to reconstruct the incoming direction of UHE cosmic rays. By training th e SBI algorithm using realistic simulations produced with CoREAS and ZHAir eS\, which include electric field amplitudes\, antenna positions\, and tri gger times\, we demonstrate how our algorithm learns the posterior probabi lity of the Bayesian model given the data. This approach enables us to acc ess robust error estimates in the reconstructed direction. Additionally\, we show that\, in contrast to standard "black box" machine learning method s\, our SBI technique allows us to evaluate the statistical rigor of our r esults.\n\nhttps://events.icecube.wisc.edu/event/243/contributions/10682/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10682/ END:VEVENT BEGIN:VEVENT SUMMARY:A graph neural network reconstruction for the IceAct telescopes DTSTART:20250129T210000Z DTEND:20250129T213000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10670@events.icecube.wisc.edu DESCRIPTION:Speakers: Larissa Paul (South Dakota School of Mines & Technol ogy)\n\nSituated at the geographic South Pole\, the imaging air shower tel escopes IceAct observe the atmosphere above the IceCube Neutrino Observato ry. Therefore\, the IceAct telescopes measure the electromagnetic air-show er development complementary to the air shower at the surface with IceTop and the high-energetic muonic component measured by the in-ice detector. C urrently\, three IceAct telescopes are installed and have shown to operate in the harsh conditions at the South Pole successfully. The telescope cam era consists of 61 silicon photomultipliers (SiPMs) with a hexagonal light guide glued to each SiPM. A graph neural network is used to reconstruct t he air-shower properties of the camera images. The graph gives great flexi bility to do a combined analysis of several telescopes and other detector components. The current status of the reconstruction method is presented.\ n\nhttps://events.icecube.wisc.edu/event/243/contributions/10670/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10670/ END:VEVENT BEGIN:VEVENT SUMMARY:Improving Gamma-ray Angular Resolution with Convolutional Neural N etwork De-noiser DTSTART:20250130T142000Z DTEND:20250130T144000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10684@events.icecube.wisc.edu DESCRIPTION:Speakers: Ruo-Yu Shang (Barnard College\, Columbia University) \n\nImaging Atmospheric Cherenkov Telescopes (IACT) reconstruct the locati ons of gamma-ray sources using stereo analysis of images of gamma-ray air showers. The images of gamma-ray showers suffer from the noise fluctuation arises from night-sky brightness. Understanding the quality of an image i s crucial for estimating the uncertainty of the gamma-ray arrival directio n. In this presentation\, we show how to improve the gamma-ray angular res olution by denoising the gamma-ray shower images with a Convolutional Neur al Network (CNN) and estimating the direction uncertainty of a gamma-ray e vent by propagating the location uncertainties of single photoelectrons in the gamma-ray camera frame.\n\nhttps://events.icecube.wisc.edu/event/243/ contributions/10684/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10684/ END:VEVENT BEGIN:VEVENT SUMMARY:A Hybrid Approach to Event Reconstruction for Atmospheric Cherenko v Telescopes Combining Machine Learning and Likelihood Fitting (Remote) DTSTART:20250128T164500Z DTEND:20250128T171500Z DTSTAMP:20260521T143700Z UID:indico-contribution-10690@events.icecube.wisc.edu DESCRIPTION:Speakers: Jim Hinton (Max-Planck-Institut für Kernphysik)\, G eorg Schwefer (Max-Planck-Institut für Kernphysik)\, Robert Parsons (Inst itut für Physik\, Humboldt-Universität zu Berlin)\n\nThe imaging atmosph eric Cherenkov technique currently provides the highest angular resolution achievable in astronomy at very high energies. High resolution measuremen ts provide the key to progress on many of the key questions in high energy astrophysics. The huge potential of the next generation Cherenkov Telesco pe Array Observatory (CTAO) in this regard can be realised with the help o f improved algorithms for the reconstruction of the air-shower direction a nd energy. Hybrid methods combining maximum-likelihood fitting techniques with neural networks represent a particularly promising approach.\nHere\, we present the FreePACT algorithm\, a hybrid machine-learning likelihood r econstruction method for IACTs. In this\, the analytical likelihood used i n traditional image-likelihood fitting techniques is replaced by a neural network that approximates the charge probability density function for each pixel in the camera. The performance of this improved algorithm is demons trated using simulations of the planned CTAO Southern array.\n\nhttps://ev ents.icecube.wisc.edu/event/243/contributions/10690/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10690/ END:VEVENT BEGIN:VEVENT SUMMARY:Modeling IACT Gamma-ray Background using Singular Value Decomposit ion DTSTART:20250130T144000Z DTEND:20250130T150500Z DTSTAMP:20260521T143700Z UID:indico-contribution-10685@events.icecube.wisc.edu DESCRIPTION:Speakers: Ruo-Yu Shang (Barnard College\, Columbia University) \n\nExtended $\\gamma$-ray sources\, such as TeV halos\, evolved pulsar wi nd nebulae\, and star clusters\, impose a challenge to analyses of Imaging Atmospheric Cherenkov Telescope (IACT) data due to the difficulty in esti mating irreducible background originating from cosmic-ray-induced $\\gamma $-ray-like events in the source regions. A background estimation method is necessary to address IACT analyses in the cases when the source angular s ize exceeds or occupies a significant part of the field-of-view. The propo sed new method analyzes the distribution of cosmic-ray-like events in the coordinate of $\\gamma$-ray camera using singular value decomposition (SVD ) to derive the irreducible background estimation. This data-driven method significantly reduces the systematic uncertainty on the background estima tion\, and the method performance is evaluated using VERITAS archival obse rvations.\n\nhttps://events.icecube.wisc.edu/event/243/contributions/10685 / LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10685/ END:VEVENT BEGIN:VEVENT SUMMARY:Machine Learning Techniques for Neutrino Reconstructions in IceCub e DTSTART:20250128T190000Z DTEND:20250128T194500Z DTSTAMP:20260521T143700Z UID:indico-contribution-10677@events.icecube.wisc.edu DESCRIPTION:Speakers: Philip Weigel (Massachusetts Institute of Technology )\n\nAdvancements in machine learning have improved event reconstruction i n the analyses of IceCube data\, providing fast and accurate estimations o f neutrino properties. These methods typically use pulse series data and t he spatial information of digital optical modules as inputs to neural netw orks. I will discuss current reconstruction techniques in IceCube and thei r applications to physics analyses. Increasingly more complex models like graph neural networks and transformers are being explored as improvements over convolutional neural network-based reconstructions. I will introduce state space models as a promising approach for efficiently reconstructing very long data sequences\, removing the need for compression techniques an d avoiding the quadratic complexity of attention. Additionally\, efforts a re ongoing to integrate these new models into the open-source reconstructi on framework GraphNeT\, which could extend these techniques to other exper iments.\n\nhttps://events.icecube.wisc.edu/event/243/contributions/10677/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10677/ END:VEVENT BEGIN:VEVENT SUMMARY:Interpretable Deep Learning for Event Reconstruction in IceCube DTSTART:20250128T194500Z DTEND:20250128T203000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10691@events.icecube.wisc.edu DESCRIPTION:Speakers: Mirco Huennefeld (Universität Dortmund)\n\nEvent re construction is a critical step in the analysis of data at the IceCube Neu trino Observatory. Traditional maximum-likelihood methods\, while provably optimal under certain conditions\, can be computationally expensive and i nfeasible in practice. A reconstruction method is presented that combines the statistical rigor of maximum-likelihood estimation with the powerful r epresentation learning capabilities of deep neural networks. By leveraging domain knowledge and exploiting inherent symmetries in the problem\, a hi ghly interpretable deep learning model is developed that improves event re construction accuracy and computational efficiency. The model not only ach ieves state-of-the-art performance but also provides robust generalization along built-in symmetries.\n\nhttps://events.icecube.wisc.edu/event/243/c ontributions/10691/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10691/ END:VEVENT BEGIN:VEVENT SUMMARY:Stereograph: stereoscopic event reconstruction using graph neural networks applied to CTAO (Remote) DTSTART:20250130T150500Z DTEND:20250130T153000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10688@events.icecube.wisc.edu DESCRIPTION:Speakers: Thomas Vuillaume (LAPP\, Univ. Savoie Mont-Blanc\, C NRS)\, Hana Ali Messaoud (LAPP\, Univ. Savoie Mont-Blanc\, CNRS)\, Tom Fra ncois (LAPP\, Univ. Savoie Mont-Blanc\, CNRS)\n\nThe CTAO (Cherenkov Teles cope Array Observatory) is an international observatory currently under co nstruction. With more than sixty telescopes\, it will eventually be the la rgest and most sensitive ground-based gamma-ray observatory.\n\nCTAO studi es the high-energy universe by observing gamma rays emitted by violent phe nomena (supernovae\, black hole environments\, etc.). These gamma rays pro duce an atmospheric shower upon entering the atmosphere\, which emits fain t blue light\, observed by CTAO’s highly sensitive cameras. The event re construction consists of analyzing the images produced by the telescopes t o retrieve the physical properties of the incident particle (mainly direct ion\, energy\, and type).\n\nA standard method for performing this reconst ruction consists of combining traditional image parameter calculations wit h machine learning algorithms\, such as random forests\, to estimate the p article's energy and class for each telescope. A second step\, called ster eoscopy\, combines these monoscopic reconstructions into a global one usin g engineered weighted averages.\n\nIn this work\, we explore the possibili ty of using Graph Neural Networks (GNNs) as a suitable solution for combin ing information from each telescope. The "graph" approach aims to link obs ervations from different telescopes\, allowing analysis of the shower from multiple angles and producing a stereoscopic reconstruction of the events . We apply GNNs to CTAO-simulated data from the Northern hemisphere and sh ow that they are a very promising approach to improving event reconstructi on\, providing a more performant stereoscopic reconstruction. In particula r\, we observe better energy and angular resolutions and enhanced separati on between gamma photons and protons compared to the Random Forest method. \n\nhttps://events.icecube.wisc.edu/event/243/contributions/10688/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10688/ END:VEVENT BEGIN:VEVENT SUMMARY:Machine learning using NuDot DTSTART:20250130T213500Z DTEND:20250130T220500Z DTSTAMP:20260521T143700Z UID:indico-contribution-10697@events.icecube.wisc.edu DESCRIPTION:Speakers: Masooma Sarfraz\, Spencer Axani (University of Delaw are)\n\nNuDot is a ton-scale liquid scintillator research and development testbed. It aims to develop techniques to reduce one of the dominant backg rounds in large modern and future liquid scintillator neutrinoless double beta decay (0νββ) searches: the solar neutrino background. With the hel p of machine learning and high-speed electronics\, NuDot will demonstrate the ability to extract directional information by separating the prompt Ch erenkov radiation within the isotropic scintillation emission. This separa tion is done using low time-transit-spread photomultiplier tubes. We are u sing U-Net architecture\, a convolutional neural network originally develo ped to perform image segmentation that aims to find the hit time of the ph oton and extract the charge. In addition\, efforts are underway to integra te these machine learning models into the front-end data acquisition syste ms\, such as RFSoC platforms\, to enable real-time processing and decision -making.\n\nhttps://events.icecube.wisc.edu/event/243/contributions/10697/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10697/ END:VEVENT BEGIN:VEVENT SUMMARY:Evaluation of energy reconstruction performance of the Telescope A rray surface detector using a deep neural network and hybrid data (Remote ) DTSTART:20250129T143000Z DTEND:20250129T150000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10693@events.icecube.wisc.edu DESCRIPTION:Speakers: Anton Prosekin (Institute of Physics\, Academia Sin ica)\, Kozo Fujisue (Institute of Physics\, Academia Sinica)\, Anatoli Fe dynitch (Institute of Physics\, Academia Sinica)\, Hiroyuki Sagawa (Insti tute for Cosmic Ray Research\, the University of Tokyo)\n\nAccurate recons truction of Ultra-High-Energy Cosmic Ray (UHECR) properties is crucial for studying their origins and composition. In this work\, we introduce a Dee p Neural Network (DNN) model based on the AixNet architecture to reconstru ct UHECR parameters using data from the Telescope Array surface detector ( SD). The DNN predicts key parameters\, such as energy\, arrival direction\ , core position\, Xmax\, and primary mass\, by analyzing both time traces and spatial correlations in the data. Monte Carlo simulations for four mas s groups (proton\, helium\, CNO\, and iron) indicate that the DNN enhances the resolution of energy\, direction\, and core position compared to stan dard methods. This improvement is expected to hold even with relaxed data quality criteria\, potentially increasing the number of usable events. We present resolution estimates\, systematic studies based on simulations\, a nd validate the DNN’s performance with hybrid data.\n\nhttps://events.ic ecube.wisc.edu/event/243/contributions/10693/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10693/ END:VEVENT BEGIN:VEVENT SUMMARY:Deep Learning in Astroparticle Physics DTSTART:20250128T160000Z DTEND:20250128T164500Z DTSTAMP:20260521T143700Z UID:indico-contribution-10694@events.icecube.wisc.edu DESCRIPTION:Speakers: Jonas Glombitza (RWTH AACHEN UNIVERSITY)\n\nDeep Lea rning in Astroparticle Physics\n\nAlgorithms based on machine learning hav e been extraordinarily successful across many domains\, including computer vision\, machine translation\, engineering\, and science.\nMoreover\, in the field of physics\, the importance of machine learning is growing quick ly\, driven by the need for precise and efficient algorithms that can effe ctively handle vast amounts of complex and high-dimensional data.\nRecentl y\, with the help of these novel algorithms\, providing improved reconstru ctions\, new insights into astroparticle physics could be gained.\nCould i t even become a new paradigm for data-driven knowledge discovery?\n\nIn th is review\, we explore the current state of machine learning in astroparti cle physics after introducing its fundamental concepts.\nWe outline the im mense potential of this emerging technology\, illustrate the wide variety of possible applications in the context of astroparticle physics\, and deb ate the latest breakthroughs.\nFinally\, we present novel approaches and t echniques and discuss future applications and challenges in the field.\n\n https://events.icecube.wisc.edu/event/243/contributions/10694/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10694/ END:VEVENT BEGIN:VEVENT SUMMARY:A Simulation-Based Inference Method for Electric Field Reconstruct ion (Remote) DTSTART:20250129T194000Z DTEND:20250129T195500Z DTSTAMP:20260521T143700Z UID:indico-contribution-10683@events.icecube.wisc.edu DESCRIPTION:Speakers: Thomas McKinley (SFSU)\, Emily Weissling\, Oscar Mac ias (San Francisco State University)\n\nThe primary goal of the Giant Radi o Array for Neutrino Detection (GRAND) is to uncover the mysterious source s of ultra-high-energy cosmic rays (UHECRs). GRAND aims to achieve this by detecting electric fields generated by UHECR interactions with Earth's at mosphere and magnetic field. Reconstructing the electric field from measur ed antenna voltages is difficult due to the need for a detailed model of t he antenna's response and background noise. \nWe will present a simulation -based inference model trained to learn the likelihood ratio using realist ic simulations from CoREAS and ZHAireS. The model incorporates a realistic antenna response and Galactic background noise to accurately reconstruct the electric field. Additionally\, we will introduce various statistical t ests\, such as coverage tests\, to demonstrate the statistical validity of our findings.\n\nhttps://events.icecube.wisc.edu/event/243/contributions/ 10683/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10683/ END:VEVENT BEGIN:VEVENT SUMMARY:Machine Learning at Telescope Array (Remote) DTSTART:20250129T140000Z DTEND:20250129T143000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10678@events.icecube.wisc.edu DESCRIPTION:Speakers: Ivan Kharuk (Institute for Nuclear Research RAS)\n\n Telescope Array is a large-scale cosmic-ray observatory studying ultra-hig h-energy cosmic rays. Its Surface Detector array consists of 507 scintilla tion stations arranged in a rectangular grid covering approximately 700 km ². This talk presents our deep learning approach to reconstructing cosmic ray properties from Telescope Array Surface Detector data. We demonstrate how combining multiple data representations with various neural architect ures (convolutional\, recurrent\, and transformer networks) enhances recon struction accuracy of primary particle properties\, including arrival dire ction and energy. Finally\, we present post-processing techniques develope d for searching for rare event\, such as ultra-high-energy photons.\n\nhtt ps://events.icecube.wisc.edu/event/243/contributions/10678/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10678/ END:VEVENT BEGIN:VEVENT SUMMARY:UHE Cosmic Ray Candidate Identification in RNO-G Deep Antennas Usi ng Machine Learning DTSTART:20250129T190000Z DTEND:20250129T192000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10696@events.icecube.wisc.edu DESCRIPTION:Speakers: Bryan Hendricks (Pennsylvania State University)\n\nU ltra-high-energy cosmic rays (UHECRs) are believed to originate from the u niverse's most cataclysmic events\, yet their sources remain unidentified. Composed primarily of protons and nuclei ranging from light elements to i ron\, these charged particle emissions are deflected en route to Earth by magnetic fields\, obscuring their true source directions. The Radio Neutri no Observatory in Greenland (RNO-G) addresses this issue by targeting UHE neutrinos\, as many of the mechanisms hypothesized to produce UHECRs are a lso expected to unleash UHE neutrinos. Neutrinos\, due to their neutrality \, near-zero mass\, and weak interactions\, can traverse the cosmos unhind ered and undeflected. While these characteristics make neutrinos invaluabl e for tracing their origins\, they also make them incredibly difficult to detect.\n\nRNO-G overcomes this challenge by leveraging the Askaryan effec t: when UHE neutrinos interact within a dense\, dielectric medium\, they p roduce showers that emit broadband electromagnetic radiation which coheren tly sums in the radio regime. This phenomenon allows for a massive effecti ve detection volume due to the long attenuation lengths of radio waves in ice. However\, UHECR showers can produce impulsive emissions that closely mimic neutrino-induced showers\, making them a critical background to acco unt for in neutrino searches. This work presents methodology and prelimina ry results from a linear discriminant analysis\, along with plans for othe r classification methods\, applied to a subset of RNO-G data to identify c osmic ray candidate events in its deep in-ice antennas.\n\nhttps://events. icecube.wisc.edu/event/243/contributions/10696/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10696/ END:VEVENT BEGIN:VEVENT SUMMARY:Reconstruction of energy and arrival directions of UHECRs register ed by fluorescence telescopes with a neural network (Remote) DTSTART:20250129T150000Z DTEND:20250129T153000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10679@events.icecube.wisc.edu DESCRIPTION:Speakers: Mikhail Zotov (Lomonosov Moscow State University)\n\ nFluorescence telescopes are important instruments widely used in modern e xperiments for registering ultraviolet radiation from extensive air shower s (EASs) generated by cosmic rays of ultra-high energies. We present a pro of-of-concept convolutional neural network aimed at reconstruction of ener gy and arrival directions of primary particles using model data for two te lescopes developed by the international JEM-EUSO collaboration. We also de monstrate how a simple convolutional encoder-decoder can be used for EAS t rack recognition. The approach is generic and can be adopted for other flu orescence telescopes.\n\nhttps://events.icecube.wisc.edu/event/243/contrib utions/10679/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10679/ END:VEVENT BEGIN:VEVENT SUMMARY:IceTop gamma-hadron separation and angular error estimation using machine learning techniques DTSTART:20250130T192000Z DTEND:20250130T194000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10673@events.icecube.wisc.edu DESCRIPTION:Speakers: Sebastian Vergara Carrasco\n\nThe IceCube Neutrino O bservatory\, located at the South Pole\, combines two detector systems to study high-energy cosmic-ray events. The surface array\, called IceTop\, i ndirectly detects cosmic rays within the 100 TeV to EeV range through ice- Cherenkov tanks\, providing reconstructed observables such as primary ener gy and direction. The in-ice optical array detects high-energy muonic comp onents of air showers. Together\, these detectors could enhance particle-t ype discrimination\, though at the cost of a narrower field of view for so urce searches. This work in progress aims to differentiate between photon- and cosmic-ray-induced air showers detected by IceTop. This is done using a Convolutional Neural Network (CNN) that processes time\, charge\, and l ateral distance distributions. The resulting classification results are th en compared against previous methods incorporating data from both detector s. Furthermore\, to support gamma-ray source searches\, we apply a boosted decision tree for estimating directional reconstruction errors. These tre e-based models excel in regression and classification tasks\, where we use numerous reconstruction fit parameters as inputs to obtain the angular er ror on an event-by-event basis.\n\nhttps://events.icecube.wisc.edu/event/2 43/contributions/10673/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10673/ END:VEVENT BEGIN:VEVENT SUMMARY:Direction and energy reconstruction with uncertainty quantificatio n for GRAND using graph neural network (Remote) DTSTART:20250129T164000Z DTEND:20250129T170000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10668@events.icecube.wisc.edu DESCRIPTION:Speakers: Arsène Ferrière (CEA-List)\, for the GRAND Collabo ration\, Aurélien Benoit-Lévy (CEA-List)\n\nFor experiments such as GRAN D\, a distributed radio-antenna array for ultra-high-energy neutrino detec tion\, a precise direction and energy reconstruction is essential. Machine -learning methods and in particular graph neural networks (GNN) appear to be an interesting solution given their ability to use localised and variab le-size inputs. In this contribution\, we will present and summarize the o ngoing work within the GRAND collaboration using GNNs in reconstruction ef forts\, and show that by adding physical inputs to our networks\, we achie ved better accuracy than existing maximum likelihood methods without incre asing the inference time. We also implemented methods to estimate the unce rtainty of our predictions under certain conditions.\n\nhttps://events.ice cube.wisc.edu/event/243/contributions/10668/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10668/ END:VEVENT BEGIN:VEVENT SUMMARY:IceTop-CNN: Cosmic-Ray Reconstruction in IceTop using a Convolutio nal Neural Network with Low-Level Inputs (Remote) DTSTART:20250130T194000Z DTEND:20250130T200000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10692@events.icecube.wisc.edu DESCRIPTION:Speakers: Ethan Dorr\, Frank McNally (Mercer University)\n\nWe present on the development of an application for training and evaluating convolutional neural networks for use with high-statistics\, minimally-cut cosmic-ray anisotropy studies. This application has been built to utilize computing resources from the IceCube Observatory using the HTCondor workl oad management system. Our goal is to streamline the creation of lightweig ht models that can successfully reconstruct well-captured and uncontained events over a large zenith range using only low-level charge and time info rmation as inputs. By doing this\, we aim to minimize systematic uncertain ty in our models while maintaining the accuracy of models trained on highe r-level parameters. Our current baseline model is capable of estimating th e energies of 68% of unfiltered simulations in our dataset within 15% of t heir true values. The application is intended to be accessible to novices in machine learning and data science with guides for installation and crea ting and assessing models available. We hope to see improvement in both th e reconstructions of additional event characteristics and accessibility of our application to beginners in programming and research.\n\nhttps://even ts.icecube.wisc.edu/event/243/contributions/10692/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10692/ END:VEVENT BEGIN:VEVENT SUMMARY:AI Agents for Ground-Based Gamma Astronomy (Remote) DTSTART:20250130T163000Z DTEND:20250130T170000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10675@events.icecube.wisc.edu DESCRIPTION:Speakers: Sergo Golovachev (JetBrains)\, Dmitriy Kostunin (DES Y)\, Vladimir Sotnikov (JetBrains)\, Strube Alexandre (Forschungszentrum J ülich)\n\nThe next generation instruments for ground-based gamma-ray astr onomy are marked by a substantial increase in complexity with dozens of te lescopes. This leap in scale introduces significant challenges in managing system operations and offline data analysis. The methods\, which depend o n advanced personnel training and sophisticated software\, become increasi ngly strained as the system's complexity grows\, making it more challengin g to effectively support users in such a multifaceted environment.\n\nTo a ddress these challenges\, we propose the development of AI agents based on instruction-finetuned large language models (LLMs). These agents align wi th specific documentation and codebases\, understand the environmental con text\, operate with external APIs\, and communicate with humans in natural language. Leveraging the advanced capabilities of modern LLMs\, which can process and retain vast amounts of information\, these AI agents offer a transformative approach to system management and data analysis by automati ng complex tasks and providing intelligent assistance.\n\nWe present two p rototypes aimed at integrating with the Cherenkov Telescope Array Observat ory pipelines for operations and offline data analysis. The first prototyp e automates data model implementation and maintenance for the Configuratio n Database of the Array Control and Data Acquisition (ACADA). The second p rototype is an open-access code generation application tailored for data a nalysis based on the Gammapy framework.\n\nhttps://events.icecube.wisc.edu /event/243/contributions/10675/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10675/ END:VEVENT BEGIN:VEVENT SUMMARY:Searching for Rare Astrophysical Events with Rare AI DTSTART:20250130T205000Z DTEND:20250130T213500Z DTSTAMP:20260521T143700Z UID:indico-contribution-10704@events.icecube.wisc.edu DESCRIPTION:Speakers: Aobo Li\n\nRare event searches are fundamental to ou r understanding of crucial astrophysical phenomena\, including neutrinoles s double beta decay\, dark matter detection\, and binary black hole merger s. While artificial intelligence has revolutionized many scientific fields \, its application to rare event searches presents unique challenges due t o the inherent scarcity of training data. This talk presents two innovativ e AI solutions specifically developed for rare event searches in physics a nd astronomy. First\, we introduce a Rare Event Surrogate Model\, initiall y designed for optimizing neutrinoless double-beta decay detectors\, with planned extensions to binary black hole merger simulations. Second\, we di scuss our AI-ready data release from a cutting-edge axion dark matter dete ctor\, demonstrating significant improvements in dark matter search sensit ivity through AI-driven analysis. These developments showcase how carefull y tailored AI approaches can overcome the challenges of limited data avail ability while enhancing our capability to detect and analyze rare astrophy sical events.\n\nhttps://events.icecube.wisc.edu/event/243/contributions/1 0704/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10704/ END:VEVENT BEGIN:VEVENT SUMMARY:Detection of Radio Signals from Cosmic Rays Using Convolutional Ne ural Networks with Data from SKALA antennas at IceTop DTSTART:20250129T162000Z DTEND:20250129T164000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10695@events.icecube.wisc.edu DESCRIPTION:Speakers: Abdul Rehman (University of Delaware)\, Paula Gálve z Molina\, Frank Schroeder (University of Delaware / Karlsruhe Institute o f Technology)\n\nCosmic rays colliding with atmospheric particles produce cascades of secondary particles known as extensive air showers. These show ers emit electromagnetic radiation whose radio component is detectable by radio antennas. At the surface of the IceCube Neutrino Observatory in Anta rctica\, a prototype station equipped with three antennas has been collect ing background and air-shower data since 2020. Traditionally\, we have emp loyed Signal-to-Noise Ratio (SNR) cuts to select candidate radio events\, which discarded valuable measurements of air showers at lower SNR levels. However\, Convolutional Neural Networks (CNNs) can outperform traditional methods in classification and denoising radio pulses from air showers. Suc h CNNs have been trained on the waveforms resulting from combining South P ole background data with simulated cosmic-ray signals generated by the CoR EAS software.The CNNs can identify additional air-shower events that do no t pass traditional SNR cuts\, while also improving the accuracy of pulse p ower and timing measurements. Recently\, we have also started to explored the impact of upsampled waveforms on the accuracy of CNN-based classifiers and denoisers. These networks will contribute to achieving the science go als envisioned for radio arrays for air-shower detection\, such as the Ice Cube-Gen2 Surface Array.\n\nhttps://events.icecube.wisc.edu/event/243/cont ributions/10695/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10695/ END:VEVENT BEGIN:VEVENT SUMMARY:Fast Generation of Realistic Data-Driven Stereoscopic Shower Image s using Generative Adversarial Networks (Remote) DTSTART:20250130T160000Z DTEND:20250130T163000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10686@events.icecube.wisc.edu DESCRIPTION:Speakers: Samuel Spencer (Friedrich-Alexander-Universität)\, Lucy Fortson (University of Minnesota)\, Hugh Dickinson (Open University)\ , Ramanakumar Sankar (University of California\, Berkeley)\, Deivid Ribeir o (University of Minnesota)\, Kameswara Bharadwaj Mantha (University of Mi nnesota)\n\nEffective identification and characterization of particle show ers captured by ground-based cherenkov telescopes is critical for very hig h energy gamma-ray astrophysics. A common step employed in the field is t o use synthetic gamma-ray and hadronic events generated based on computati onally-expensive simulations and use them for downstream analyses. Leverag ing the power of generative deep learning\, various studies have developed fast emulators that can generate synthetic simulated events that mimic th e simulation outputs. However\, they still carry the intrinsic assumption that the emulated/simulated data is representative of the observations. In an attempt to address the aforementioned challenges\, in this work\, we d esigned and trained on real data a fully-unsupervised Wasserstein Generati ve Adversarial Network on stereoscopic shower images (Stereo-wGAN) from th e VERITAS gamma-ray observatory. In this presentation\, we highlight our m odel’s ability to generate realistic stereoscopic events that are self-c onsistent in their quantitative image-wise moments (Hillas parameters) and overall reconstructed shower parameters. We also showcase the utility of our model-learnt internal feature representations in exploring the diversi ty of shower events as a way towards enabling future unsupervised characte rization of gamma-ray and hadronic events\, which is another challenging t ask in the field of gamma-ray astrophysics.\n\nhttps://events.icecube.wisc .edu/event/243/contributions/10686/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10686/ END:VEVENT BEGIN:VEVENT SUMMARY:Convolutional Neural Network Processing of Radio Emission for Nucl ear Composition Classification of Ultra-High-Energy Cosmic Rays (Remote) DTSTART:20250129T171500Z DTEND:20250129T173000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10676@events.icecube.wisc.edu DESCRIPTION:Speakers: Paula Gina Isar (Institute of Space Science — INFL PR Subsidiary\, 077125 Bucharest-Magurele\, Romania)\, Cosmina Mihoreanu ( Faculty of Automatic Control and Computer Science\, National University of Science and Technology Politehnica Bucharest\, 060042 Bucharest\, Romania )\, Emil Ioan Slusanschi (Faculty of Automatic Control and Computer Scien ce\, National University of Science and Technology Politehnica Bucharest\, 060042 Bucharest\, Romania)\, Tudor Alexandru Calafeteanu (Faculty of Au tomatic Control and Computer Science\, National University of Science and Technology Politehnica Bucharest\, 060042 Bucharest\, Romania)\n\nUltra-hi gh-energy cosmic rays (UHECRs) are the most mysterious particles in the Un iverse originating from extragalactic sources\, which yet rise a couple of fundamental open questions\, such as where do they come from\, how do the y propagate\, and how do they reach the energies they exhibit. Due to the very low flux\, i.e. one particle per km2 per century at about 1020 eV\, U HECRs are detected indirectly from the Earth\, through their developed air showers\, by modern detection techniques in frame of hybrid and large-sca le experiments. Radio detectors have proven to be a competitive method for reconstructing the properties of EASs\, such as the shower’s incoming d irection\, its energy\, and its maximum development (Xmax). \nConcurrently \, data science has become indispensable in physics. By applying statistic al\, computational\, and deep learning methods to large databases\, resear chers can extract insights and make predictions efficiently and accurately \, in conjunction with traditional analysis methods. \nWe introduce a conv olutional neural network (CNN) architecture designed to classify simulated CoREAS air shower events to process the radio emission for several types of primary cosmic rays’ nuclei. For the classification of the primary pa rticle\, we use metrics like Accuracy or MCC to indicate the prediction ca pability for mass-composition based on data that can be gathered by the Ra dio Detector (RD) at the world’s largest cosmic ray experiment on Earth\ , the Pierre Auger Observatory.\n\nhttps://events.icecube.wisc.edu/event/2 43/contributions/10676/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10676/ END:VEVENT BEGIN:VEVENT SUMMARY:In-situ pulser depth reconstruction for RNO-G using Neural Network DTSTART:20250129T192000Z DTEND:20250129T194000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10680@events.icecube.wisc.edu DESCRIPTION:Speakers: David Besson (University of Kansas)\, Sanyukta Agarw al (University of Kansas)\n\nRadio Neutrino Observatory in Greenland (RNO- G) aims to detect Askaryan emission from ultra-high energy astrophysical a nd cosmogenic neutrinos above 10 PeV. Situated at Summit Station\, it is p roposed to have 35 stations of which 7 stations have been installed so far . Search for neutrinos and their direction reconstruction using interferom etry requires precise control of parameters such as antenna positions and an accurate ice model. Various known sources are available in and around t he RNO-G stations which can be used in calibration of the observatory. In- situ calibration pulsers deployed on helper strings in each station along with pulser drops performed for some stations allow us to constrain the un certainty in antenna position and test the accuracy of our ice model. \n\n In my poster I assume a simple straight line\, plane wave approximation an d ignore ray-bending as an initial guess to reconstruct the depth of the s tationary pulsers in 14 helper strings across 7 stations. The station geom etry allows this simple model to be a good approximation and I use the sta tionary pulser data to train my neural network\, allowing me to reconstruc t pulser depths in cases where ray-bending effects might be more significa nt (such as pulser drops). This method is preferred as it’s much faster than analytical raytracting or simulating radio propagation.\n\nhttps://ev ents.icecube.wisc.edu/event/243/contributions/10680/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10680/ END:VEVENT BEGIN:VEVENT SUMMARY:Machine learning-based analyses using surface detector data of the Pierre Auger Observatory DTSTART:20250128T214500Z DTEND:20250128T223000Z DTSTAMP:20260521T143700Z UID:indico-contribution-10672@events.icecube.wisc.edu DESCRIPTION:Speakers: Pierre Auger Collaboration\, Steffen Traugott Hahn ( KIT - IAP/ETP)\n\nThe Pierre Auger Observatory\, the world’s largest det ector for studying ultra-high-energy cosmic rays (UHECRs)\, employs multip le detection techniques to observe the different components of extensive a ir showers. In order to accurately understand the physics of UHECRs\, it i s essential to determine their mass composition. Since UHECRs can only be measured indirectly\, it is necessary to study mass-sensitive observables\ , such as the number of muons reaching the ground and the atmospheric dept h of the shower maximum. One way to estimate these observables is by analy zing spatio-temporal patterns in the shower footprint recorded by the surf ace detector (SD) of the Observatory. Given the complexity of this informa tion\, the Pierre Auger Collaboration utilizes machine learning (ML) to co mplement the traditional analytical techniques. With the SD operating near ly 100% of the time\, ML algorithms enable the analysis of events with an unprecedented precision. In this work\, we summarize the ML-driven analyse s conducted at the Pierre Auger Observatory to identify and reconstruct ma ss-sensitive observables and explores potential applications of ML in othe r areas. Special emphasis is placed on techniques that utilize the differe nt sub-detector systems of the SD\, including the newly installed scintill ator detectors from AugerPrime\, highlighting their potential for advancin g UHECR studies.\n\nhttps://events.icecube.wisc.edu/event/243/contribution s/10672/ LOCATION:Clayton Hall (University of Delaware) URL:https://events.icecube.wisc.edu/event/243/contributions/10672/ END:VEVENT END:VCALENDAR