Rebuild tests

[mmaelicke]

The old tests are a bit messy. We should change them all over:

• move the tests folder to the top level
• get rid of unittest and move to pytest
• test against the working Matlab code
• build a github action to run tests

@AlexDo1 do you want to take on this? Otherwise assign to me.

[AlexDo1]

I emailed Uwe to see if we can meet tomorrow so he can give me some explanations about his Matlab code.

So I can do it. I'm also interested in trying out Github actions and some Matlab practice wouldn't be bad either.

[AlexDo1]

I am currently working on the tests.
What would you use as test data?
Some 'random' numpy data as in the current tests or should I use some data from metacatalog, for example?

[mmaelicke]

I think for information theory it is perfectly fine to create random samples. In scipy and sklearn are also methods to create samples from joint probabilities and clustered data.

[AlexDo1]

All right, I would then go by how you did it in the old test with np.random.seed and two distributions.
I would then use the two data sets shown here for testing:

[mmaelicke]

perfect.

[mmaelicke]

You can save the data to txt and calculate test results with Uwe's code