Loading timeline…
19971990s
Long Short-Term Memory (LSTM)
An RNN architecture capable of learning long-term dependencies.
Why It Was Important
Standard RNNs suffered from the 'vanishing gradient problem', meaning they forgot early inputs in long sequences. Sepp Hochreiter and Jürgen Schmidhuber solved this by inventing the LSTM cell, incorporating 'gates' to selectively remember or forget data. LSTMs dominated sequence modeling (translation, speech) until Transformers arrived.
Who Invented It
Sepp Hochreiter & Jürgen Schmidhuber
German researchers pushing the boundaries of recurrent neural networks.
Applications
- Speech Recognition
- Machine Translation
- Time Series Forecasting
Key Papers
- Long Short-Term Memory
Sepp Hochreiter, Jürgen Schmidhuber · Neural Computation · 1997
Videos
Long Short-Term Memory (LSTM), Clearly Explained
StatQuest with Josh Starmer
What is LSTM (Long Short Term Memory)?
IBM Technology