.. _method-multilevel_function_train-regression_type:

"""""""""""""""
regression_type
"""""""""""""""


Type of solver for forming function train approximations by regression


.. toctree::
   :hidden:
   :maxdepth: 1

   method-multilevel_function_train-regression_type-ls
   method-multilevel_function_train-regression_type-rls2


**Specification**

- *Alias:* None

- *Arguments:* None


**Child Keywords:**

+-------------------------+--------------------+--------------------+---------------------------------------------+
| Required/Optional       | Description of     | Dakota Keyword     | Dakota Keyword Description                  |
|                         | Group              |                    |                                             |
+=========================+====================+====================+=============================================+
| Required (Choose One)   | Regression Type    | `ls`__             | Use least squares solver for forming        |
|                         |                    |                    | function train approximations by regression |
|                         |                    +--------------------+---------------------------------------------+
|                         |                    | `rls2`__           | Use regularized regression solver for       |
|                         |                    |                    | forming function train approximations       |
+-------------------------+--------------------+--------------------+---------------------------------------------+

.. __: method-multilevel_function_train-regression_type-ls.html
__ method-multilevel_function_train-regression_type-rls2.html



**Description**


Function train approximations are formed based on regression for the set of
coefficients described at :dakkw:`model-surrogate-global-function_train`.  Solver
options include least squares and regularized regression, where the latter
penalizes high-order terms to mitigate overfitting.

*Default Behavior*

The default regression solver is least squares.