A white dwarf is what remains of a star like the Sun after it sheds its outer layers. What's left is a hot core the size of Earth but with a mass close to that of the Sun. A teaspoon of this material would weigh as much as a truck. Some of these stars have extremely strong magnetic fields — a billion times stronger than Earth's. But such objects are rare in catalogs: they are dim, and typical sky surveys miss them.
Scientists took data on several thousand white dwarfs with hydrogen atmospheres and applied two machine learning algorithms. The first, dimensionality reduction, compresses complex data into an intuitive map where similar objects end up close together. The second, clustering, looks for natural groupings on that map. This revealed distinct groups of white dwarfs, each with its own mass, temperature, and age.
The most interesting result: by knowing what already confirmed magnetic white dwarfs look like, the algorithm estimated the field strength for hundreds of other stars that had no direct measurements. It turned out that such 'hidden magnets' are far more numerous in the sample than previously thought. Machine learning acted like a smart filter that finds rare objects among ordinary ones.
This approach is like sorting fruit on a conveyor belt: by color, size, and density, the machine separates rare varieties that a human might miss. Only here, instead of apples, we have stars, and instead of a conveyor belt, we have telescope data.
A white dwarf's magnetic field can be a billion times stronger than Earth's, such a field would tear apart atoms near the star's surface.