Physics inside AI: scientists created a neural network that understands the laws of crystallography
American researchers have developed a physics-aware neural network called DONUT that learns on its own from the laws of X-ray interaction with matter and decodes the nanostructure of materials in real time. The breakthrough is reported in the journal npj Computational Materials.
Researchers in the United States have developed a 'physics-savvy' neural network called DONUT, which independently learns from the laws of X-ray interaction with matter and deciphers the nanostructure of materials in real time. The study was published in the journal npj Computational Materials. The work was reported by Hightech.fm. The network is designed to understand crystallography principles, enabling it to interpret X-ray diffraction patterns without extensive labeled training data. This approach promises faster and more accurate analysis of material structures, which could accelerate research in materials science and related fields.
Source: Hightech.fm —
original