For example, I trained a cnn model on five of mnist digits (0,1,2,3,4) and 
now I am trying to add another class (5) on already trained model. I don't 
want to use the data of (0,1,2,3,4) classes again, I just want to update 
the model with the data of the new class (5). Is it possible?

What I have done is that:
I defined a 10-class cnn model but I give only  (0,1,2,3,4) digits to train 
it. The testing a digit from  (0,1,2,3,4) classes gave plausible results. I 
saved the model
Later I load the model as by 'get_value()  -  set_value()'. Then I give a 
dataset that only contains class(5) data. But in final, my model params 
didn't update itself as I was expected.

Thanks in advance,

-- 

--- 
You received this message because you are subscribed to the Google Groups 
"theano-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to