Types of RNNs

Although in the RNN structure shown previously, $T_x = T_y$ i.e. the lengths of input and output sequences were the same, this is not always the case.

One-to-One RNN

Input and output sequences are of unit length.

One-to-Many RNN

Input sequence is of unit length but the output sequence has a length greater than 1.

Ex. Music Generation, input x could be the genre

Many-to-One RNN

Input sequence has a length greater than 1, but the output sequence is of unit length.

Ex. Text Sentiment Classification

Many-to-Many RNN

Input and output sequences are of equal length

Ex. Named Entity Recognition

Input and output sequences are of different lengths

Ex. Machine Translation i.e. language translation, say from English to French

PreviousRNN Structure NextLanguage Modeling

Last updated 3 years ago

Although in the RNN structure shown previously, $T_x = T_y$ i.e. the lengths of input and output sequences were the same, this is not always the case.