Skip to content
BENIBIX Learn

Recurrent Neural Network (RNN)

AI📈 Machine LearningNeural Networks

Also known as RNN

A network that reads in order. It keeps a short memory of what came before.

Think of it like

Think of it like reading a story one word at a time. You keep the plot in mind as you go.

Example

Jack's phone keyboard guesses his next word. An RNN reads 'see you' and, with that in memory, offers 'tomorrow'.

Why it matters

It was the main tool for language and speech for years. Today transformers do most of that work.

Definition

A recurrent neural network reads data in order, one piece at a time. After each piece, it keeps a small memory and passes it to the next step. This lets it use what came before, such as the earlier words in a sentence. It was widely used for speech and text. Its memory fades over long stretches, and later designs tried to fix that.
Each box reads one word and hands its memory to the next box along.

Common mistake

Expecting it to remember a whole page. Its memory fades, so it often loses track of words from long ago.

Key takeaways

  1. 1
    In order. One piece after another.
  2. 2
    Carries memory. From each step to the next.
  3. 3
    Memory fades. Long stretches are hard.

Learn it properly