A TT-eigenvalue solver that finally works
26/05/2016
In a recent preprint by Maxim Rakhuba and Ivan Oseledets we propose a new algorithm for calculation of vibrational spectra of molecules using tensor train decomposition. Under the assumption that eigenfunctions lie on a low-parametric manifold of low-rank tensors we introduce the concept of manifold preconditioning of well-known iterative methods (inverse iteration and locally optimal block conjugate gradient method). As an application, we accurately compute vibrational spectra (84 states) of acetonitrile molecule on a laptop in one hour using only 180 MB of memory to represent all computed eigenfunctions.
