GitLab

Along with @mbertoni , we have realized that the implementation of classes such as proj1, clus1, etc in their methods fit and predict only consider as input a signature instance. After starting to create notebooks, I have realized that sometimes our input for those fit and predict will be data queried from other signature that will be in a form of numpy array or a csv file. Currently, the only way is to put the data in h5 file and to mock a signature with that h5 file. We think we should find a better way. There are several possibilities to deal with that.

1. Accept a signature or and .h5 file as inputs. The h5 file should contain the datasets "V" and "keys".
2. Move the implementation of fit and predict to private methods which do not care about the different possible inputs. Then, we could create different methods for the different types of input data. Those methods would parse the input data and call the private method.

Maybe, other options could be possible but we should discuss it, since the final solution will affect several classes.