Closed sigveka closed 9 years ago
@SurrogateModel.py:199: OK @SurrogateModel.py:703: OK
Shouldn't the bounds that are passed in the data-structure? ie.
parameters={
'param0': { 'min': 0.0, 'max': 10.0, 'argPos': 0 },
'param1': { 'min': 0.0, 'max': 10.0, 'argPos': 1 },
},
Please also update the Fortran version of twoTank to support autoInit.
Sorry for the late reply I did not notice.
That is correct, it was my bad.
I honestly do not understand the "autoInit" example (it does not work, for one it lacks cmake files for the exact task). What are you trying to do?
I am ready with more code to implement initialisation from input output data. I am a little lost on gitHub right now, will those changes be included in this pull request? (it says 700+ lines are changed)
The changes are somewhat comprehensive, especially to surrogate function.
New strategies (in Strategy.py
)
InitialData
, lines: 287-325NonLinFitToPointWithSmallestError
, lines: 535 - 669InitialDataPoints
, lines: 782-795Changes to surrogate models in (SurrogateModel.py
):
ForwardMappingModel
:
*
updateFitDataFromFwSpec
identical to the method inBackwardMappingModel
, lines: 494-510
*updateMinMax
identical to the method inBackwardMappingModel
, lines: 512-539
*error
identical to the method inBackwardMappingModel
, lines: 533, 552
*parameterFittingStrategy
, lines 568-569
* lines: 482-489, Check wether “initial strategy” is given as an argument.
BackwardMappingModel
:
*
updateMinMax
Hi Sigve, it is good practice to stay on the topic both in conversations and branches. It makes it much easier to follow and merge. This requires submitting self-contained patches and opening new conversations when the topic changes. I know this can be difficult at times.
For your questions:
Henrik
Good morning/evening,
I'm sorry about that. I thought I had to handle pull requests on the gitHub web-page, and there was no button to create separate pull requests there.
Sigve
On 4 June 2015 at 08:23, Henrik Rusche notifications@github.com wrote:
Hi Sigve, it is good practice to stay on the topic both in conversations and branches. It makes it much easier to follow and merge. This requires submitting self-contained patches and opening new conversations when the topic changes. I know this can be difficult at times.
For your questions:
-
The manual pull request suggests git checkout -b sigveka-modelModules modelModules git pull https://github.com/sigveka/MoDeNa.git modelModules This currently updates to 625120 "Added new strategy for initialisation...". This should have been separate pull request because it closes #7 https://github.com/MoDeNa-EUProject/MoDeNa/issues/7
autoInit: Yes, CMakeLists.txt was missing. Added & pushed. It works like this: an exception is throw if a model is not initialised before the workflow is started. The exception is catched and call the initialisation within the workflow.
Henrik
— Reply to this email directly or view it on GitHub https://github.com/MoDeNa-EUProject/MoDeNa/pull/9#issuecomment-108748032 .
You can probably only have one (open) pull request per branch. It would be good if they only contain a single topic because that makes it easier to review them. Please consider using "git rebase" to manufacture a clean history.
I was getting several errors since the last push.
Note fixes in line 199 and 703 of SurrogateModel.py respectively saving the surrogate function in the database and a call to cModel.
I am also planning to add the parameter bounds to the fitting in the call to
nlfb(...)
in line 429 in Strategy.py. However, I am uncertain wether it is smart to add the upper and lower bounds (line 252 253 in SurrogateModel.py) should be aFloatField
inSurrogateModel
orSurrogateFunction
(I prefer the later).