And by the way, I made it easier to solve quantum physics problems, but scientists don't yet understand how

And by the way, I made it easier to solve quantum physics problems, but scientists don't yet underst

Machine learning and artificial intelligence are capable of making breakthroughs in many applications and scientific fields. The most interesting promises are the advances in quantum physics. What happens in the quantum world is hard to understand from a common sense point of view, but in terms of mathematics, there is nothing unexplained, although it is still difficult to solve quantum equations. The new approach promises to teach the AI how to solve these problems much faster.

For understandable reasons, simple models are used to describe quantum phenomena in mathematics and physics, but even then there are hundreds, thousands and even millions of equations to describe the interaction processes of a limited number of particles at the level of quantum mechanics. For example, to visualize the interaction model of two electrons in the crystallal grid node, 100,000 equations are needed, one per pixel of visualization. This requires enormous computing resources.

An international group of Italian and American physicists and mathematicians were able to create a model of machine learning that reduced the problem to just four equations per pixel. And without losing it exactly. True, it took two weeks of intensive math training, but the result worked. Moreover, the proposed model could be used to solve other problems for an active mathematical machine, the method of renormalization, which would expand the scope of application of the proposed tool by physics elementary particles and neuroscience.

" said Domenico Di Santa, the lead author of the study.

The renormalization group method usually operates many parameters and supports the scaling of processes. Researchers have created an AI model that first creates connections in a full-scale renormalization group without simplification, and then modifys these relationships to bring all calculations to a small set of equations with a similar result. The final result is retained, but the path to it is different by many orders of computational power. One problem, scientists do not yet understand how the EI calculates ways of optimization, but they intend to understand this in future research.