Bulletin of the American Physical Society
APS April Meeting 2021
Volume 66, Number 5
Saturday–Tuesday, April 17–20, 2021; Virtual; Time Zone: Central Daylight Time, USA
Session L17: Data Science IILive
|
Hide Abstracts |
Sponsoring Units: GDS Chair: Dimitri Bourilkov, University of Florida |
Sunday, April 18, 2021 3:45PM - 3:57PM Live |
L17.00001: Inference of neutrino flavor evolution through data assimilation and neural differential equations Ermal Rrapaj, Amol Patwardhan, Eve Armstrong, George Fuller The evolution of neutrino flavor in dense environments such as core-collapse supernovae and binary compact object mergers constitutes an important and unsolved problem. Its solution has potential implications for the dynamics and heavy-element nucleosynthesis in these environments. In this paper, we build upon recent work to explore inference-based techniques for the estimation of model parameters and neutrino flavor evolution histories. We combine data assimilation, ordinary differential equation solvers, and neural networks to craft an inference approach tailored for non-linear dynamical systems. Using this architecture, and a simple two-neutrino-beam, two-flavor model, we compare the performances of nine different optimization algorithms and expand upon previous assessments of the efficacy of inference for tackling problems in flavor evolution. We find that employing this new architecture, together with evolutionary optimization algorithms, accurately captures flavor histories in the small-scale model and allows us to quickly explore both model parameters and initial flavor content. In future work we plan to extend these inference techniques to large numbers of neutrinos. [Preview Abstract] |
Sunday, April 18, 2021 3:57PM - 4:09PM Live |
L17.00002: Improving IceCube Event Reconstruction Using A Graph Convolutional Network And Semantic Segmentation Rui An Machine learning is a candidate for the next-generation reconstruction for neutrino experiments such as IceCube. IceCube is an ice-Cherenkov neutrino detector embedded in a cubic kilometer of glacial ice in Antarctica. The detector observes astrophysical and atmospheric neutrinos via the light emitted by charged particles produced in neutrino interactions with 5160 digital optical modules~(DOM). A typical $\nu_\mu$ interaction~(1TeV$\sim$100TeV) originating inside the detector, namely the ``starting track'', is dominated by deep inelastic scattering which produces a hadronic cascade near the interaction vertex and a muon track. Conventional reconstruction assumes continuous energy loss for ``starting track'' events in the hadronic cascades, leading to a generally underestimated energy reconstruction. Correct clustering the hadronic cascade and track is crucial for improving the energy reconstruction. This study presents a graph convolutional network for semantic segmentation to distinguish the DOM charges as cascade-like and track-like charges. [Preview Abstract] |
Sunday, April 18, 2021 4:09PM - 4:21PM Live |
L17.00003: Fast Energy Reconstruction using CNNs for GeV Scale Neutrinos in IceCube Jessie Micallef The IceCube Neutrino Observatory, located deep under the Antarctic ice, detects astrophysical and atmospheric neutrinos. It uses 5160 optical modules spanning a cubic kilometer of ice to detect Cherenkov radiation originating from neutrino interactions. Atmospheric neutrinos at the scale of 10-GeV can be used to measure important fundamental properties of neutrinos such as the oscillation parameters and to search for non-standard interactions. Current likelihood-based reconstructions take seconds to minutes to reconstruct the properties (energy, direction, etc.) of a neutrino event, which makes them computationally challenging for large data sets. In this talk, I will present work showing the optimization of a convolutional neural network (CNN) to reconstruct the energy of 10-GeV scale events in IceCube. This method takes sub-milliseconds per neutrino event, and also offers improvements in the energy resolution. [Preview Abstract] |
Sunday, April 18, 2021 4:21PM - 4:33PM Live |
L17.00004: Evolving Antennas for Ultra-High Energy Neutrino Detection Alexander Patton Evolutionary algorithms borrow from biology the concepts of mutation and selection to approach complex problems more efficiently than traditional techniques. The GENETIS project uses genetic algorithms to develop antenna designs with higher sensitivity to radio impulses from ultra-high energy neutrino interactions than current designs. We attempt to improve on antenna designs used in current experiments, using geometric constraints imposed by a narrow hole in deep ice. By integrating the XFdtd finite-difference time domain modeling program with simulations of neutrino experiments, we are able to assign a fitness score that is based on neutrino sensitivities. We will report on advancements to the algorithm, steps taken to improve the software we use, the latest results from our evolutions, as well as our roadmap for manufacturing. [Preview Abstract] |
Sunday, April 18, 2021 4:33PM - 4:45PM Live |
L17.00005: 1D to 2D CYLEX Transformation to Save Time in JWL Parameterization Reid Ginoza A Jones-Wilkins-Lee (JWL) equation of state (EOS) can be parameterized by simulating a cylinder expansion experiment (CYLEX) and selecting those parameters that lead to the closest match of the experimental velocity history data. This may be computationally expensive as it may take many simulations to find an acceptable parameter set. In this study, we investigate the possibility of using a hypothetical one-dimensional sphere expansion simulation as a substitute for the two-dimensional CYLEX simulation. The goal was to find a transform function that maps the velocity history of the sphere simulation to the velocity history of the CYLEX simulation. JWLs from several sources, some randomly generated, were used in both the one-dimensional sphere expansion and the two-dimensional CYLEX simulations. Three regression models were then constructed and evaluated: a three-term polynomial regression, a nine-term polynomial regression, and a multivariate adaptive regression spline model. These regression models resulted in close agreement between the CYLEX simulation velocity history and the transformed one-dimensional sphere expansion simulation velocity history. [Preview Abstract] |
Sunday, April 18, 2021 4:45PM - 4:57PM On Demand |
L17.00006: Combined research approach for determining the ground state of spin glass Alena Korol, Dmitrii Kapitan, Alexey Rybin, Egor Vasiliev, Konstantin Soldatov, Yuriy Shevchenko, Vitalii Kapitan, Konstantin Nefedev Monte Carlo simulation is one of the most powerful approaches in statistical physics. However, the ability of modern machine learning techniques to classify, identify, or interpret massive data sets provides a complementary paradigm to the above approach. In our research, we studied the Edwards-Anderson (EA) spin glass (SG) model in a square two-dimensional (2D) lattice of Ising spins with a bimodal distribution of bonds. In present work, for simulation we used a combination of our Hybrid Multispin Method (HMM) and the Restricted Boltzmann Machine (RBM) to predict the ground states for the Edwards--Anderson spin glass model. To predict the GS, we used the data of the HMM to train our neural network and to predict spin glass state with a lower energy level than in the training data sets. Our research has shown that the ground states of spin glass systems can be predicted using a neural network. However, given the absence of an exact solution to determine the number of degenerations at the ground energy level for big systems, it is impossible to check whether our algorithms reach the global energy minimum. Nonetheless, machine-learning algorithms can make a significant contribution to solving NP-complete problems in the future. [Preview Abstract] |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700