Neutrino Fingerprints: Image-Based Encodings of IceCube Events for CNN Direction Reconstruction

Name: Neutrino Fingerprints: Image-Based Encodings of IceCube Events for CNN Direction Reconstruction
Uploaded: 2026-10-21T21:08:57.100507+00:00
Duration: 743 s
Description: The work introduces neutrino fingerprints, compact 72x72x3 images derived from IceCube event data to reconstruct incoming neutrino directions using convolutional neural networks. By encoding detector timing, charge, and their difference into RGB channels, the method provides an interpretable baseline for direction reconstruction.

The work introduces neutrino fingerprints, compact 72x72x3 images derived from IceCube event data to reconstruct incoming neutrino directions using convolutional neural networks. By encoding detector timing, charge, and their difference into RGB channels, the method provides an interpretable baseline for direction reconstruction.

Open the full episode