Currently the .save method fails since the rosettascio writer can't deal with the SmoothNMF object in self.learning_results.decomposition_algorithm.
We need a save method that saves all the information obtained from the espm decomposition if it has already been performed.
It is not clear to me right now if the best solution is to write a custom save method for the EDS_espm class or to bring up the issue to rossettascio. In my EELSNMF project I have custom load/save function which makes things easier, with the only inconvience is that you then would use espm.load instead of hs.load.
Another option is to store (at the very least) G,W,H in the metadata manually.
Can you check that you also have this issue and give feedback on what do you think is the best way to adress it?
Currently the .save method fails since the rosettascio writer can't deal with the SmoothNMF object in self.learning_results.decomposition_algorithm.
We need a save method that saves all the information obtained from the espm decomposition if it has already been performed.
It is not clear to me right now if the best solution is to write a custom save method for the EDS_espm class or to bring up the issue to rossettascio. In my EELSNMF project I have custom load/save function which makes things easier, with the only inconvience is that you then would use espm.load instead of hs.load.
Another option is to store (at the very least) G,W,H in the metadata manually.
Can you check that you also have this issue and give feedback on what do you think is the best way to adress it?