Scientific Lego: Predicting Material Properties Without a Lab ⚡ экспресс
Original: "Materials Graph Library (MatGL), an open-source graph deep learning library for materials science and chemistry"
The free MatGL program acts like a virtual tester: from the arrangement of atoms, it predicts a substance's characteristics in hours, not months.
Links in the knowledge graph 1
Atoms are like Lego blocks. In the past, to understand a material, you had to experiment for a long time. Now, the MatGL library lets you quickly 'play out' different combinations of atoms on a computer, predicting properties. This will speed up the creation of new batteries and solar panels.
🎯 The graph neural networks underlying MatGL are also used in social networks: 'friend' algorithms analyze connections between people in the same way as between atoms.
🎬 The idea of instant material selection is reminiscent of replicators from 'Star Trek', where matter is synthesized according to a given program.