Regular grid interpolation #244
Conversation
|
Hi @flabowski. What about adding a preprocessing step within this class to create the 1d vectors from the parameters given by the |
…nts, not the grid axes.
|
Hi @mtezzele |
There was a problem hiding this comment.
Hi @flabowski, I left some comments on your code. Let me know if you need any clarification. Thanks!
|
I realized now that one of the changes @fandreuz suggested changed the behaviour and one of the tests failed. I assume that was unintended. I reverted it, so the behaviour matches the |
ezyrb/regular_grid.py
Outdated
| :return: the interpolated values. | ||
| :rtype: numpy.ndarray | ||
| """ | ||
| new_point = np.asarray(new_point) | ||
| if len(new_point.shape) == 1: | ||
| new_point.shape = (-1, 1) |
|
Dear @flabowski, thanks for the effort! From my side, everything looks almost perfect, I just asked if you can add a simple test in order to check that the exception is raised in case of not grid points. |
docs/source/regular_grid.rst
Outdated
| @@ -10,10 +10,10 @@ RegularGrid | |||
| :nosignatures: | |||
|
|
|||
| RegularGrid | |||
| RegularGrid.def | |||
There was a problem hiding this comment.
This is not needed. Or at least we don't use it in the other files. I am not sure about the generated outcome...
There was a problem hiding this comment.
I am not sure either. The file was generated automatically when ran make_rst.sh. Some other things seem to be fishy.
edit: everything seems fine after reverting what make_rst.sh did. @ndem0 maybe some commands need to be updated?
There was a problem hiding this comment.
We are not using that script anymore! Sorry, it's not written, I gonna remove it in the next release of EZyRB (#232).
The file now looks correct, thanks!
* added tests for regular grid interpolation * added regular grid interpolation module * added the rst file for the regular grid interpolation
see #243
The class needs a grid, i.e. something like
grid = ([0.1, 0.2, 0.3], [1, 2, 3, 4, 5]), not a list of points such aspoints = [[0.1, 1], [0.1, 2], [0.1, 3], ..., [0.3, 5]]. That means rom.fit() does not work, the method assumes we want to use the points from the database to construct the interpolator. Instead the user has to callreduction.fit(...)first and thenrg_approximation.fit(...)with the proper grid.Although the interpolation returns what we expect in the tests, I have not tested it in a full ROM yet.