Add Features: absolute_maximum; mean_n_absolut_max | Add install_requirements.sh script

[OliEfr]

Added:

F1. I added a install_requirements.sh script for convenience
F2. I added a absolut_maximum feature. This feature calculates the absolut maximum.
F3. I added a mean_n_absolut_max feature. This calculates the arithmetic mean of the n absolut maximum values of the time series.

Tests:

T1: I added unit-tests for mean_n_absolut_max.
T2: I did not ad unit-tests for absolut_maximum, because there is also no unit-test for maximum.

T3: Some tests fail. Most are not on my watch, I think. However, I think the following are on me but I dont know how to resolve them: test_contains_all_non_high_comp_features and test_default_calculates_all_features. In particular, I dont know how I can add my newly created features here.

Questions:

Q1: The feature mean_n_absolut_max takes an additional input parameter (the number of maximums to consider). Where and how should I check, that this input parameter is valid, e.g. not smaller 0 and so on. Should it be checked in the pytest unittests? Or should I raise an error within the feature calculator function itself? For now, I just return np.nan if the input parameter is not valid.

Others

• build the docs
• Sorry for commiting multiple changes at once and sorry for not creating a new branch. I think, I should have done this. Next time I will.

[MaxBenChrist]

good first draft, btw. what are those lock files in the dask-worker space for?

[OliEfr]

I noticed that you dont test for the AssertionError in test_estimate_friedrich_coefficients. Is there a reason for this?

I added assertRaises tests for test_mean_n_absolute_max.

[OliEfr]

@MaxBenChrist Could you please refer to T3 from my PR? What about test_contains_all_non_high_comp_features and test_default_calculates_all_features?

[nils-braun]

Also thanks from my side. I had a few comments in the code.

Some general points:

• as mentioned by Max, please remove the dask-worker-space files from your PR.
• one of your new features (the mean_n_absolute_max) requires a parameter. This means you need to add it to the feature calculator settings with a reasonable default. Especially, you need to add an entry in the ComprehensiveFCParameters dict (have a look into the other features which are already there). I would assume reasonable values would be small numbers, probably below 10.

To your questions:

• Q1: testing for < 0 in your code is fine. Users can change the parameters, so testing in the unittests is not enough.
• no, there is no reason we do not test the assertion error. That could be added, good point.
• to T3: To fix the two remaining tests, you need to add your new feature to the settings (see above) - as mentioned in the exception of the tests.

[OliEfr]

Check. All done.

For testing the install_requirements.sh file, I created another issue. Because I am not sure yet how to implement it.
I understand, that it makes sense to include it in the workflow. However, no workflow so far requires this exact install_requirements.sh file.

[nils-braun]

I will have a look this evening, but I think I am still seeing the dask worker files :-)
Thanks already for your time!

[nils-braun]

Thanks!