Sepp Hochreiter

Papers

Total Citations

H-Index

About

Sepp Hochreiter is a pioneering force in deep learning, best known for co-inventing the Long Short-Term Memory (LSTM) architecture, a foundational breakthrough that revolutionized sequence modeling. His research spans neural networks, meta-learning, and artificial intelligence, with a focus on making machines capable of learning how to learn. In his seminal work "Meta-learning with backpropagation" (2002, 64 citations), Hochreiter introduced gradient-based methods for meta-learning, enabling neural networks to adapt rapidly to new tasks—a concept now central to modern AI agents, robotics, and non-stationary time series analysis. His more recent contributions include "Trusted Artificial Intelligence: Towards Certification of Machine Learning Applications" (2021, 24 citations), addressing the critical need for reliable and certifiable AI systems in society. Most recently, his work "A Large Recurrent Action Model: xLSTM enables Fast Inference for Robotics Tasks" (2024) extends LSTM principles to large-scale reinforcement learning, offering efficient alternatives to Transformer-based models. With over 100,000 citations across his career, Hochreiter’s LSTM remains one of the most cited inventions in computer science, underpinning advances in speech recognition, language translation, and autonomous systems. His legacy continues to shape AI research, from foundational theory to trusted, real-world applications.