SLAC's SPINE AI tool decodes elusive neutrino data

Researchers at SLAC National Accelerator Laboratory have developed SPINE, a machine-learning tool that reconstructs neutrino interactions from detector data. On real data from Fermilab's ICARUS detector it recovered 60% more usable interactions than conventional methods. SPINE is set to be deployed on eight detectors, including DUNE.

• SPINE recovered 60% more usable neutrino interactions on ICARUS data
• The tool combines multiple ML models instead of a chain of narrow algorithms
• SPINE is already trained for DUNE detectors, expected to start by 2030
• Deployment is planned across eight detectors