Logo image
Back
Neural posterior estimation of the neutrino direction in IceCube using transformer-encoded normalizing flows on the sphere
Preprint   Open access

Neural posterior estimation of the neutrino direction in IceCube using transformer-encoded normalizing flows on the sphere

R Abbasi, M Ackermann, J Adams, J. A Aguilar, M Ahlers, J. M Alameddine, S Ali, N. M Amin, K Andeen, C Argüelles, …
ArXiv.org
21 Apr 2026
DOI: https://doi.org/10.48550/arxiv.2604.19846
url
https://doi.org/10.48550/arXiv.2604.19846View
Preprint (Author's original) Open arXiv.org - Non-exclusive license to distribute

Abstract

Computer Science - Artificial Intelligence Computer Science - Learning Physics - High Energy Astrophysical Phenomena Physics - High Energy Physics - Experiment Physics - Instrumentation and Methods for Astrophysics
IceCube is a cubic-kilometer-scale neutrino detector located at the geographic South Pole. A precise directional reconstruction of IceCube neutrinos is vital for associations with astronomical objects. In this context, we discuss neural posterior estimation of the neutrino direction via a transformer encoder that maps to a normalizing flow on the 2-sphere. It achieves a new state-of-the-art angular resolution for the two main event morphologies in IceCube - tracks and showers - while being significantly faster than traditional B-spline-based likelihood reconstructions. All-sky scans can be performed within seconds rather than hours, and take constant computation time, regardless of whether the posterior extent is arc-minutes or spans the whole sky. We utilize a combination ofC² -smooth rational-quadratic splines, scale transformations and rotations to define a novel spherical normalizing-flow distribution whose parameters are predicted as a whole as the output of the transformer encoder. We test several structural choices diverting from the vanilla transformer architecture. In particular, we find dual residual streams, nonlinear QKV projection and a separate class token with its own cross-attention processing to boost test-time performance. The angular resolution for both showers and tracks improves substantially over the whole trained energy range from 100 GeV to 100 PeV. At 100 TeV deposited energy, for example, the median angular resolution improves by a factor of1.3for throughgoing tracks, by a factor of1.7for showers and by a factor of2.5for starting tracks compared to state-of-the art likelihood reconstructions based on B-splines. While previous machine-learning (ML) efforts have managed to obtain competitive shower resolutions, this is the first time an ML-based method outperforms likelihood-based muon reconstructions above 100 GeV.

Metrics

1 Record Views

Details

Logo image