Sequential Models

Abstract

So why do we need a recurrent neural network (RNN)? Let’s try to answer that with an example, or analogy. When reading a new article, people have two options. First, if they can’t understand it, they can read articles that the new article is based on, for background information. Otherwise, they do understand the new article, based on some prior knowledge of the subject, without an immediate need to read similar articles. In both cases, their ability to understand the new article is enabled by some preexisting or preaquired knowledge. But they don't need to go back to the phase of learning alphabets or numbers; they only need to know what this article is about. This is the way recurrent neural networks work.