Currently the data processing and loading is coupled to the model, which violates the single responsibility principle, and makes debugging hard as data has to be reloaded when only the model is getting debugged, and vice versa. There should be a data layer that is responsible for handling the data, separate from the model.
Arguably the plotting functions have the same issue, there should be another function or object that takes in the model and data as input, then plots the learning curves.
Currently the data processing and loading is coupled to the model, which violates the single responsibility principle, and makes debugging hard as data has to be reloaded when only the model is getting debugged, and vice versa. There should be a data layer that is responsible for handling the data, separate from the model.