Browse Teach Curate

Sequence-to-Sequence Learning

An encoder LSTM compresses the input sequence into a final state that initializes a decoder LSTM, which emits the output sequence one token at a time through a softmax over the vocabulary. The single fixed-size bottleneck vector is its defining weakness on long inputs.

jeremy Video

Sequence-to-sequence Encoder-Decoder Neural Networks using LSTMs for English to Spanish Translation

Sequence-to-sequence modeling utilizes a dual-stage architecture comprising distinct encoder and decoder components to map variable-length input sequences to variable-length output sequences. The the…