An Inexact Tensor-Train Primal-Dual Interior-Point Method for Semidefinite Programs

Abstract

In this work, we introduce an interior-point method that employs tensor decompositions to efficiently represent and manipulate the variables and constraints of semidefinite programs, targeting problems where the solutions may not be low-rank but admit low-tensor-train rank approximations. Our method maintains approximate superlinear convergence despite inexact computations in the tensor format and leverages a primal-dual infeasible interior-point framework. In experiments on Maximum Cut, Maximum Stable Set, and Correlation Clustering, the tensor-train interior point method handles problems up to size $2^{12}$ with duality gaps around $10^{-6}$ in approximately 1.5~h and using less than 2~GB of memory, outperforming state-of-the-art solvers on larger instances. Moreover, numerical evidence indicates that tensor-train ranks of the iterates remain moderate along the interior-point trajectory, explaining the scalability of the approach. Tensor-train interior point methods offer a promising avenue for problems that lack traditional sparsity or low-rank structure, exploiting tensor-train structures instead.