Keahi Pelkum Donahue, a recent high school graduate, has developed an innovative AI model using Transformer networks to forecast solar flares. This project earned him recognition in the Regeneron Science Talent Search and a scholarship.
Donahue's model, typically used in speech processing, was adapted to analyze solar magnetic field data. It demonstrated superior performance with limited data, matching the accuracy of models using larger datasets. Solar flares can disrupt satellites and telecommunications, making accurate predictions crucial for mitigating potential damage.
Donahue will continue his education at Princeton University, focusing on physics and vocal performance, with aspirations to earn a Ph.D. in plasma physics and contribute to nuclear fusion research.