Speaker
Dmitry Salnikov
(INR RAS & Lomonosov MSU)
Description
The numerical calculation of observables in quantum theory is reduced to the sampling from a set of random variables with a joint distribution density defined analytically on the basis of physical theory. This problem is solved by various modifications of the Metropolis method, but it requires significant computational costs with a large sample size and the number of random variables describing the physical system. An alternative approach based on generative machine learning models is being considered. The features of the problem, such as translational symmetry, are analyzed in the context of building the most optimal architecture.
Primary authors
Dmitry Salnikov
(INR RAS & Lomonosov MSU)
Vsevolod Chistiakov
(MSU, Faculty of Physics)
