A Recipe for Understanding LSTMs & GRUs

We will proceed to prove that LSTMs & GRU are easier than you thought

Although RNNs might be what first cross your mind when you hear about natural language processing or sequence models, most success in the field is not attributed to them but rather (and for a long time) to an improved version of the RNN that solves its vanishing gradient problem. In particular, an LSTM (Long Short-term Memory) network or less often, a GRU (Gated Recurrent Unit) network. …

gru lstm nlp recipe recurrent neural network understanding