juniorDeep Learning
What is a Recurrent Neural Network (RNN)?
Updated May 16, 2026
Short answer
An RNN is a neural network architecture designed for sequential data using recurrent connections.
Deep explanation
Unlike feedforward networks, RNNs maintain hidden states that preserve information from previous time steps. This allows them to model temporal dependencies in sequences such as text, speech, and time-series data. At each step, the hidden state is updated using current input and previous hidden state. Traditional RNNs suffer from vanishing gradients, which led to advanced variants like LSTM and GRU.
Real-world example
RNNs are used in speech recognition, machine translation, and stock market forecasting.
Common mistakes
- Using standard RNNs for long sequences where LSTMs are more suitable.
Follow-up questions
- Why do RNNs struggle with long-term dependencies?
- What is hidden state?
- What are sequence models?