LINEAiGE
AI HISTORY

Long Short-Term Memory

Sepp Hochreiter and Jürgen Schmidhuber published Long Short-Term Memory in Neural Computation on 15 November 1997. The paper introduced the gradient-based method called long short-term memory (LSTM), using constant error flow and multiplicative gate units to address long time-lag learning problems in recurrent networks.

15 November 1997AI HISTORYVERIFIED

Historical classification

This verified record belongs to LINEAiGE's artificial-intelligence history layer beginning in 1956.

Why it matters

A documented recurrent-neural-network milestone whose original publisher record explicitly names and describes the LSTM method and its mechanism for learning across long time intervals.

People, organizations & technologies

PEOPLE
Sepp Hochreiter, Jürgen Schmidhuber

ORGANIZATIONS
Technische Universität München, IDSIA, MIT Press, Neural Computation

TECHNOLOGIES
Long short-term memory, Recurrent neural networks, Gradient-based learning

Evidence & verification

STATE
verified

CONFIDENCE
high

LAST VERIFIED
2026-09-10T12:05:10+08:00

Authors, journal citation, publication date, DOI, author affiliations and the paper's description of LSTM are verified against the MIT Press Neural Computation record and an author-hosted copy of the original paper. PubMed independently corroborates the same publication date, authors, journal citation and DOI. No claim is made here that this record alone establishes invention geography, every precursor, later influence, or causal relationships.

Sources

SOURCE 1 · PRIMARY · CLAIM-EVIDENCE

Long Short-Term Memory

MIT Press / Neural Computation

PUBLISHED
1997-11-15

SOURCE 2 · PRIMARY · CORROBORATION

Long Short-Term Memory

Jürgen Schmidhuber / IDSIA

PUBLISHED
1997-11-15

SOURCE 3 · CORROBORATION

Long short-term memory

PubMed / U.S. National Library of Medicine

PUBLISHED
1997-11-15

Trace the chronology

Record identity

LINEAiGE ID
lstm-1997