A GRU is more like a simplification of the ideas in LSTM, rather than a building block. At a high level, it uses the hidden state as the memory of the cell (rather than a separate cell state) and it uses a single "update" gate, merging the forget and input gates. Overall it performs similarly to LSTM while being more computationally efficient (fewer matrices).
Comments
A GRU is more like a simplification of the ideas in LSTM, rather than a building block. At a high level, it uses the hidden state as the memory of the cell (rather than a separate cell state) and it uses a single "update" gate, merging the forget and input gates. Overall it performs similarly to LSTM while being more computationally efficient (fewer matrices).