A neural network architecture designed for sequential data processing, where connections create loops allowing information to persist across time steps. RNNs can process variable-length sequences and maintain memory of previous inputs, useful for language modeling and time series.
