The idea of the function was to test different sets of parameters. E.g. different datasets. So the function should take some list or interval of values/paths/data and then try the different parameters given. Now it seems like this is not the case. You can start with just making the function being able to differ just one parameter for simplicity.
However maybe it is nice to have the function you made for one of these iterations. And have a wrapper function for looping over the different sets of parameters.
Other fixes:
The file should not be under notebooks. Only files under the directory anodet is part of the package.
You don't need to save the means and inverse covariances, and then load them again. So remove the saving and loading of the parameters.
Instead of visualising the results with images, I think it's easier to start with just printing the values of ROC-AUC score, recall, and precision (look in the tests_example notebook).
It seems like the function is being called in the end of the file. This should not be there.
Name the functions something more specific than run and visual, so people get a hint of what they do.
Add docstrings to the functions.
There should not be any defaults on the data paths. It is very unlikely that other people will have the data saved in the same directory structure as you.
Add an example of how to use the function in the tests_example notebook.
The idea of the function was to test different sets of parameters. E.g. different datasets. So the function should take some list or interval of values/paths/data and then try the different parameters given. Now it seems like this is not the case. You can start with just making the function being able to differ just one parameter for simplicity.
However maybe it is nice to have the function you made for one of these iterations. And have a wrapper function for looping over the different sets of parameters.
Other fixes: