recurrent neural network (RNN)

A recurrent neural network (RNN) is a neural network architecture for sequential data. This network applies the same recurrent transformation at each time step and passes a hidden state forward, allowing information to persist over time.

Classical forms include vanilla RNNs and gated variants like LSTM and GRU, which overcome vanishing and exploding-gradient problems via gating mechanisms during backpropagation through time.

Architectures can be unidirectional or bidirectional, and encoder–decoder setups (often augmented with attention) power many sequence-to-sequence tasks such as machine translation.

While transformer models now dominate many benchmarks, RNNs continue to be useful in streaming, low-latency, or resource-constrained environments.