I'm trying to implement self-organising maps in Python using pymvpa2 SOM mapper.
It works just fine, but how do I save the resulting SOM?
h5save didn't work for me and I started wondering, if I can create an empty SOM instance and then attach a saved Kohonen layer to it? However, when I try to feed the Kohonen matrix to an empty mapper I get this error:
...
size = (15,15)
som_new = SimpleSOMMapper(size, 1900, learning_rate=0.05)
som_new.K = som_trained.K
>> AttributeError: can't set attribute
Is there a way to override this?