Deep Learning - RNN - Recurrent Neural Network - Types Of RNN Tutorial

Types Of RNN -

• Many to One
• One to Many
• Many to Many
• One to One

Many to One

- Many inputs, but One output

Input - Sequential data like sentences, characters, and time series data

Output - Non Sequential data like integer/number or scalars

Applications like Sentiment Analysis (1, 0), Rating Prediction (1, 2, 3, 4, 5), etc.

One to Many

- One input, but many output

Input - Normal Non Sequential data like numbers or image

Output - Sequential data like words or sentences.

Applications like giving images as input and generating captions for it.

Input -

Output - "A boy playing cricket"

Many to Many

- Many inputs with Many outputs

Input and Output are both Sequential data like sentences, characters, and time series data

It has two of its type-

• Same Length Many to Many - Input Sequence is equal to be same as Output Sequence

Applications like - 

Part Of Speech Tagging in NLP

Name Entity Recognition

• Variable Length Many to Many - Input Sequence is not the same as Output Sequence

Machine Translation like one language to another language e.g. Google Translate

In the below example, the input is 7 words and the output is 8 word

One to One

- One input with One output

Input and Output are both Non-sequential data like numbers or images.