In: Computer Science
Consider a simple case where we have trained a NN for recognising patterns X1 to Xn belonging to some classes, say A and B. Suppose we have a new training data Xn+1 that belongs to class C. We want this also to be trained. Will the previous training be impacted? How would it work?
So, according to you the existing classes are A and B. In this case, when you have a new training instance Xn+1 that belongs to class C, it is possible to train this instance as well, without effecting the previous trained model.
Your pre-trained network has a layer, which handles the recognition of 2 original classes. The easiest (and working) trick to introduce the new class, is to use all the layers before the last as granted and add an additional layer (in a new model, or as a parallel one).
Without access to original training data, you would have two options: