Hands-On One-shot Learning with Python

Generative Modeling-Based Methods

When humans make inferences about unseen data, they make use of strong prior knowledge (or inductive bias) about related events they've seen, heard, touched, or experienced. For example, an infant who has grown up with a dog may see a cat for the first time and immediately infer that it shares similarities with the pet-like temperament of the household dog. Of course, cats and dogs as species and individuals are wildly different; however, it's fair to say that a cat is more similar to a dog than other random things the child has experienced—such as food. Humans, as opposed to machine learning models, don't need thousands of examples of cats to learn that category from scratch once they have already learned ...