1997

Sepp Hochreiter

b. 1967

Long Short-Term Memory

With Jürgen Schmidhuber, introduced LSTM to help recurrent neural networks learn dependencies over long sequences.

Their joint 1997 paper introduced a memory architecture designed to address difficulties in propagating learning signals over long intervals.

LSTMs became influential in speech recognition, translation and sequence modelling. They remain an important milestone even though Transformers later became dominant in many language applications.

Dated contributions

  1. 1997

    Long Short-Term Memory

    Sepp Hochreiter & Jürgen Schmidhuber

    With Jürgen Schmidhuber, introduced LSTM to help recurrent neural networks learn dependencies over long sequences.

    Paper · 1997 — Long Short-Term Memory ↗

Reference

Original work & historical records

Wikipedia