In this section, we're going to take it one step further and explore the question of whether a neural network could learn from other neural networks and what they've already learned. We'll start by covering the concept of transfer learning, and then we'll get into some Python code.
Transfer learning is essentially the Frankenstein's monster of machine learning. The idea arose from this question: how can I take what some other network has already learned and go from there? We're basically going to do a brain splice between several different networks. This can be extremely valuable in cases where a network is trained on data that you don't have access to or the training process is the one that would have taken hours or days, as is commonly the case in text or image processing domains.
We don't want to retrain our model because it would take forever, but we want to take what we've already learned about the other two classes and start learning something else about the...