Rename deconvolution and identification to model_estimation

[BjoernLudwigPTB]

To avoid unambiguous method naming we will combine all methods from identification and devonvolution in one new module model_estimation including the fit_filter.py methods with same namings in both modules. That requires us to rename some methods out of devonvolution.fit_filter.py. Because of the following incompatibility with previous versions of PyDynamic we need to insert a deprecation warning into the next minor release and inform users about the upcoming change.

[BjoernLudwigPTB]

• Start with deprecation warning.

[BjoernLudwigPTB]

@eichstaedtPTB What about identification/fit_transfer.py, should that be moved to the new module model_estimation as well or should we think about something else?

[eichstaedtPTB]

@eichstaedtPTB What about identification/fit_transfer.py, should that be moved to the new module model_estimation as well or should we think about something else?

This should be moved into the new module, too.

[eichstaedtPTB]

The methods that fit the reciprocal could be named "iLSFIR", "iLSIIR", etc.

[BjoernLudwigPTB]

@eichstaedtPTB What do you think about the new structure? Should we maybe have a short conversation about it or can you tell from looking at the code, that we can use this new structure for replacing identification and deconvolution?

What I did now is prefix the "old" deconvolution methods with "inv" and combine both fit_filter modules in model_estimation.fit_filter and of course transfer identification.fit_transfer into model_estimation.fit_transfer. We thought, that the proposed "i" as a prefix of "L..." would be difficult to percieve on first glance, which might lead to irritation.

I prepared the docs as well and would be grateful, if you could take a look at them and either comment or straight away contribute in the according PR #131 .

[BjoernLudwigPTB]

Introducing these changes we noticed, that we got those invLS[I,F]IR with and without uncertainties and Monte Carlo methods. We should merge these implementations and introduce optional parameters to choose one or the other.

This applies to

• invLS[F,I]IR
• invLS[F,I]IR_unc
• invLSFIR_uncMC

There is no invLSIIR_uncMC since the uncertainties are determined using MC already in invLSIIR_unc .

We should then for half a year after releasing 2.0.0 backport bug fixes to the 1.4 minor releases before we stop supporting these. The docs for 2.0.0 should be enriched with the information on how to tell pip to upgrade only to the newest version of the 1.4 minor release series.

• Besides that we actually do not fit a second-order system but rather a second-order model, thus we should rename fit_transfer.fit_sos() to fit_transfer.fit_som(). (See Rename fit_sos() to fit_som() #145 for this)

We have opened another issue for the remaining tasks. #147 .

[BjoernLudwigPTB]

What remains here is the deletion of the old modules along with the actual deprecation in v2.0.0.