What is the best way to adapt a neural network after its initial training?
I.e. I want to do some image recognition and the network should get better the more new pictures I present it. That could be done with reinforced learning but for a fast progress at the beginning I want to use back propagation. Is it possible to update a network?
And what about creating new categories later on?
Is there another way than retraining it with the complete dataset since that would take a lot of time.
Sorry for my basic questions but I couldn't find much information about this.
Neural networks can be adapted by training them with small learning rates on the new data. Maybe even training the last layers with a larger learning rate than the others (incase you are using a deep neural network).
For the second part of your question, about creating new categories, a deep neural network can be used as a feature extractor on top of any other classifier (maybe another small neural network). When you want to add a new category, you have to re-train the small classifier (or neural network). This would mean that you will retain the training values of the feature detector (the deep neural network) and use it to detect new categories.