MLFlow for Experiments

[jacobnzw]

MLFlow provides a systematic way of keeping track of experiments and results.

Would be great to set up the MLFlow interface for the models and filtering algorithms.

The organization of the experiments is shit now and has to be changed:

• The process by which the data is simulated needs to be separated from the code for setting up the SSM and the filters.
• Each filter should be able to run independently on the same data.
• Data should be generated once to a file and then fed to each filter
• This should make it possible to use Dask for running several filters in parallel on the same data, thus saving time.