Researchers have successfully classified neutrino events using a quantum computer, reaching testing accuracies near 80% with neural projected quantum kernels and approximately 70% with quantum convolutional neural networks, according to a study published August 21, 2026, in Quantum Science and Technology.
Quantum Machine Learning in Neutrino Astronomy
The study, led by Pablo Rodriguez-Grasa from the University of the Basque Country UPV/EHU, demonstrates the feasibility of applying quantum machine learning to astronomical data analysis using current hardware. According to the published findings, the research team investigated neural projected quantum kernels (NPQK) and quantum convolutional neural networks (QCNNs) to distinguish between different types of neutrino events, which are critical for understanding rare cosmic phenomena in neutrino telescopes like IceCube.
The neural projected quantum kernel approach achieved testing accuracy near 80%, demonstrating a capacity to classify neutrino events directly on simulators and the IBM Strasbourg quantum processor. Executing this procedure straight on quantum hardware removes the necessity for exclusively simulated outcomes, confirming the approach against the present noise and constraints of quantum devices, as noted by the authors.
Moment-of-Inertia Encoding Scheme
A critical challenge in applying quantum machine learning to high-energy physics is encoding large feature spaces. Traditional methods struggle with the vast amounts of information generated by neutrino telescopes, limiting the feasibility of quantum graph neural networks.
To address this obstacle, Rodriguez-Grasa and colleagues introduced a moment-of-inertia-based encoding scheme. This preprocessing strategy is inspired by the underlying physics and geometry of the problem, designed to reduce the dimensionality of the data while preserving essential physical characteristics. According to the study, this innovative approach allowed the team to work with a manageable number of qubits, making the classification task achievable on existing quantum hardware.
Comparing NPQK and QCNN Approaches
Alongside the NPQK method, the study explored quantum convolutional neural networks, achieving approximately 70% accuracy in simulated tests across a wide energy range. The QCNN outcomes exhibit a minor decrease compared to NPQK scores, yet they still highlight how quantum machine learning can benefit neutrino astronomy.
The researchers validated their approach through simulations and by implementing it on the IBM Strasbourg quantum processor. This confirmed the robustness of the results above 1 TeV and showed close agreement between simulated and hardware performance. Differentiating muon signatures from those generated by electromagnetic and hadronic showers forms the core of the classification objective, serving as an essential step to identify the flavor makeup of incoming neutrinos. To train and evaluate their quantum algorithms, the group generated a dataset containing modeled signatures for both electron and muon neutrinos, representing a notable step forward in analyzing astrophysical information via quantum platforms.
Worth a look