Blurb::
UQ method leveraging a functional tensor train surrogate model.
Description::
Tensor train decompositions are approximations that exploit low rank structure
in an input-output mapping.  Refer to the :dakkw:`model-surrogate-global-function_train`
surrogate model for additional details.

*Usage Tips*

This method is a self-contained method alternative to the
:dakkw:`model-surrogate-global-function_train` surrogate model specification, similar
to current method specifications for polynomial chaos and stochastic collocation.
In particular, this ``function_train`` method specification directly couples with a
simulation model (optionally identified with a ``model_pointer``) and an additional
``function`` train surrogate model specification is not required as these options
have been embedded within the method specification.
Topics::

Examples::

.. code-block::

    method,
     function_train
       start_order = 2
       start_rank = 2  kick_rank = 2  max_rank = 10
       adapt_rank
       solver_tolerance    = 1e-12
       rounding_tolerance  = 1e-12
       convergence_tol     = 1e-6
       collocation_points  = 100
       samples_on_emulator = 100000
       seed = 531
       response_levels = .1 1. 50. 100. 500. 1000.
       variance_based_decomp


Theory::
Faq::
See_Also::
