In this section, we'll discuss transfer learning and how we can use it to train our model easily and more efficiently. Transfer learning is the abstraction of knowledge from the original owner, in such a way that anyone can obtain it freely. In a way, it's similar to how humans transfer knowledge between generations, or to each other.

Originally, for humans, the only way to transfer experience was by talking. As you can imagine, this was crucial for our survival. Of course, we eventually found better ways of storing knowledge, that is, through writing. In this way, we preserved the knowledge in its original form for a longer period, making it more extendable. Even nowadays, the dissemination of information is a fundamental aspect of society. Almost every mathematical theory is built on top of existing ones, which were written years ago.

Now we are in the era of digitalization. Information is disseminated much faster, reaching a significant number of people. We're talking about the internet, social media, online learning, and access to information everywhere, at any time. Regardless of its evolution, the basic concept behind knowledge transferal is the same. There's abstraction of information from the owner, and then we preserve it in a form where anyone can obtain it. First, we stored it in memory, that is, in people's brains; and then in books, that is, on paper by writing it down; and finally in the electronic format, that is, in the format where other humans can freely access and read it: