Interior point method in tensor optimal transport

Abstract

We study a tensor optimal transport (TOT) problem for $d\ge 2$ discrete measures. This is a linear programming problem on $d$-tensors. We introduces an interior point method (ipm) for $d$-TOT with a corresponding barrier function. Using a "short-step" ipm following central path within $\epsilon$ precision we estimate the number of iterations.

arXiv commentCorrected typos and added a short additional subsection, 11 pages