Machine learning is learning to hear inside the Sun
The Sun is a bell. Not a metaphor — the entire solar interior resonates with acoustic waves, trapped pressure oscillations that bounce between the surface and the core roughly every five minutes. The field that studies these oscillations is called helioseismology, and for three decades a network of ground stations has been recording every pulse. A team at the University of Sheffield and the National Solar Observatory just ran 30 years of those oscillations through three different machine learning architectures. All three converge on the same prediction: Solar Cycle 25 peaked in early 2025, and the next minimum falls around 2030–2031. The paper, published this month in Solar Physics, is one of the first to treat the Sun’s acoustic frequency shifts as a forecasting signal for the solar cycle — not just a diagnostic one. ...