Researchers Classify Neutrino Events at Nearly 80% Accuracy on IBM Quantum Hardware
Updated
Updated · Quantum Zeitgeist · Aug 23
Researchers Classify Neutrino Events at Nearly 80% Accuracy on IBM Quantum Hardware
1 articles · Updated · Quantum Zeitgeist · Aug 23
Summary
Near-80% testing accuracy with a neural projected quantum kernel let researchers classify neutrino telescope events on the IBM Strasbourg quantum processor, matching traditional machine-learning performance closely enough to show current hardware can handle the task.
A moment-of-inertia encoding scheme cut the data to a qubit-manageable size while preserving key physics, addressing the large feature spaces that usually block quantum analysis of detectors such as IceCube.
About 70% accuracy from a quantum convolutional neural network in simulations across a wide energy range reinforced the result, while NPQK stayed robust above 1 TeV and tracked simulator performance closely on hardware.
The study focused on separating muon tracks from hadronic and electromagnetic cascades—a key step in determining neutrino flavor composition—and said it was the first successful quantum-computer treatment of this problem.
Published Aug. 21 in Quantum Science and Technology, the work points toward practical quantum machine-learning use in neutrino astronomy as hardware improves.