
Academic lab · Linz, Austria
Institute for Machine Learning, JKU Linz
Johannes Kepler University's machine learning institute, led by Sepp Hochreiter, the inventor of LSTM and a driver of modern Hopfield networks and xLSTM.
Linz, Austria
- Type
- Academic lab
- Based in
- Linz, Austria
- Key papers
- 4
- Featured projects
- 8
About
The Institute for Machine Learning at Johannes Kepler University Linz has been led since 2018 by Sepp Hochreiter. Hochreiter developed the long short-term memory architecture in his 1991 diploma thesis and published it with Jürgen Schmidhuber in 1997. He previously led the university's Institute of Bioinformatics and heads the LIT AI Lab, which he took over in 2017.
The group works on machine learning and statistical methods. Its researchers introduced modern Hopfield networks with continuous states in 2020 and applied them to immune repertoire classification. In 2024 the group and collaborators published xLSTM, an extended LSTM architecture presented at NeurIPS, and the company NXAI, where Hochreiter is chief scientist, builds industrial models on it.
Code and models appear on the institute's GitHub and Hugging Face organizations.
Mission and ambitionsThe institute develops and applies machine learning and statistical methods, with a stated emphasis on deep learning fundamentals and their use in science and industry.
Known for
- 01
Invention of LSTM
- 02
Modern Hopfield networks
- 03
xLSTM architecture
Milestones
- 1997
LSTM
Hochreiter and Schmidhuber published the long short-term memory architecture in Neural Computation.
- 2017
LIT AI Lab
Hochreiter became head of the Linz Institute of Technology AI Lab.
- 2018
Institute for Machine Learning
Hochreiter began leading the Institute for Machine Learning at JKU.
- 2020
Modern Hopfield networks
Ramsauer and colleagues introduced Hopfield networks with continuous states.
- 2024
xLSTM
Published Extended LSTM at NeurIPS as a scalable alternative to Transformers and state space models.
Open source
Hopfield Networks is All You Need (1,977 stars)
Modern Hopfield networks for tabular data (317 stars)
Code for Universal Physics Transformers (179 stars)
Geometry-informed neural networks (195 stars)
- ml-jku/dm_controldataset
Control task data for offline reinforcement learning.
- ml-jku/dark_roomdataset
Dark-room reinforcement learning data.
- ml-jku/mhnfsspace
Few-shot molecule property prediction with modern Hopfield networks.
Tox21 toxicity prediction leaderboard.
Research papers
Notable releases
Openness
Code is public on GitHub and datasets on Hugging Face.
People
- Sepp HochreiterHead of the institute
- Günter KlambauerResearcher, Hopfield and xLSTM co-author
- Maximilian BeckFirst author of xLSTM
- Korbinian PöppelCo-author of xLSTM
- Johannes BrandstetterCo-author of xLSTM
Listed from the lab's own pages and public profiles.
Contact and community
Contact details are given on the institute's pages of the JKU Linz website.
Sources
Compiled from the lab's public pages, Hugging Face and GitHub. Last checked 2026-11-17. Something wrong or out of date? Tell us on Discord.



