In this work, we develop an optimization framework for problems whose
solutions are well-approximated by Hierarchical Tucker (HT) tensors, an
efficient structured tensor format based on recursive subspace factorizations.
By exploiting the smooth manifold structure of these tensors, we