We introduce new methods for integrating nonlinear differential equations on low-rank manifolds. The methods are based on oblique projectors onto the tangent spaces of low-rank manifolds with an interpolation property that collocates the differential equation on the manifold. These projections yield evolution equations that enable low-rank time integration of any vectorfield that can be evaluated entry-wise. The proposed methods do not require the vectorfield to have a low-rank structure, thus overcoming significant limitations of dynamical low-rank approximation based on orthogonal projection. To construct the oblique projectors, we devise a sparse tensor sampling algorithm based on the discrete empirical interpolation method (DEIM) that parameterizes tensor train manifolds and their tangent spaces with cross interpolation. Using these projectors we propose two time integration schemes on low-rank tensor train manifolds. The first is based on a time-dependent tensor cross interpolation representation of the solution and the second is a direct generalization of the well-known orthogonal projector-splitting integrator to a oblique projectors. We demonstrate the proposed methods with applications to several tensor differential equations arising from the discretization of partial differential equations.
翻译:暂无翻译