Back to Glossary
Long Short-Term Memory (LSTM)
Models & Architectures
RNN variant with gating to prevent forgetting.
LSTMs use input, output, and forget gates to retain information over long sequences.
- Advantages: More stable gradients than vanilla RNNs.
- Applications: Speech, music, time series forecasting.
- Trade-offs: Higher computational cost than GRUs.