Sophelio Launches First Fusion Competition at NeurIPS Challenges - AI to Reconstruct Invisible Magnetic Structures Inside Fusion Reactors

In this episode of Global Economic Press, Alex Brady delves into an innovative development in fusion energy and artificial intelligence. Sophelio has launched the Fusion Equilibrium Challenge at the Neural Information Processing Systems Challenges, marking the first fusion energy competition accepted to the Neural Information Processing Systems Competition Track. This challenge, organized in collaboration with the DIII-D National Fusion Facility, the United Kingdom Atomic Energy Authority’s Fusion Analysis and Interpretation of Results program, and the University of Texas at Austin's Institute for Fusion Studies, invites the global machine-learning community to reconstruct invisible magnetic structures inside fusion reactors using only non-magnetic diagnostic measurements. This capability is crucial for future reactor-class fusion power plants, which may require such methods for routine operation due to the degradation of traditional magnetic sensors under neutron fluxes. The Fusion Equilibrium Challenge aims to determine if machine learning can reconstruct the magnetic geometry that confines a fusion plasma, using non-magnetic features like poloidal-field coil currents and Thomson-scattering electron profiles. The dataset, released under the Creative Commons Attribution 4.0 license on Hugging Face, includes over 9,100 curated plasma discharges from the DIII-D National Fusion Facility and the Mega Amp Spherical Tokamak. Participants can explore these discharges using Sophelio's Data Fusion Labeler, a free desktop platform for multimodal sensor data. The competition, which is open for registration, will culminate in December 2026 at the Neural Information Processing Systems Competition Workshop, with cash awards and co-authorship opportunities for top-performing teams. For more information, visit Sophelio's website.