In this subchapter, we explain how modern RNN such as LSTM, GRU, and Bidirectional RNNs address the problem of exploding and vanishing gradients.
