How a Computer Learns to See Nature's Invisible Patterns ✛ ✎ ✦
A new algorithm works like an artist painting on several canvases at once, each with a different level of detail: from broad brushstrokes to the tiniest dots. It creates several blurred copies of the data, then uses machine learning to find connections between them. This reveals a hidden structure that is invisible on a single layer.
The same approach, without any guidance, separated the night lights of Chengdu into districts and roads, and in images of the galaxy NGC 3627 it detected subtle dust lanes. In the interstellar medium, it picked out nebulae and pockets of star formation. Riemann once speculated that nature contains hidden connections between distant regions, and the algorithm indeed finds them. Now it's ready to analyze anything: from climate maps to the structure of materials.
The algorithm reconstructed the magnetic structure of plasma without magnetic field data — using only density distribution.
In the movie 'The Matrix', hidden codes show the true structure of reality — just as this algorithm reveals invisible connections.