Vanilla RNN

Back to Recurrent Neural Networks

The simplest recurrent architecture. Processes sequences one step at a time, maintaining a hidden state that captures information from previous steps. Suffers from the vanishing gradient problem, making it difficult to learn long-range dependencies.

  • LSTM (solves vanishing gradient problem)
  • GRU (simplified alternative to LSTM)

deep-learning rnn