Durk Kingma
Papers
1
Total Citations
87
H-Index
1
About
Durk Kingma is a leading figure in deep learning, best known for his foundational contributions to generative modeling and variational inference. He co-invented the Variational Autoencoder (VAE), a cornerstone technique that bridges deep learning with probabilistic graphical models, enabling powerful unsupervised learning of complex data distributions. His work on Adam, a widely-used stochastic optimization method, has become a standard tool in training neural networks, with tens of thousands of citations. Kingma’s research also extends to normalizing flows, as seen in his work on VideoFlow (87 citations), which applies flow-based generative models to video prediction and sequence modeling. Beyond these, he has made key advances in semi-supervised learning and representation learning, often emphasizing principled probabilistic approaches. His papers collectively command over 100,000 citations, reflecting his immense impact on the field. Kingma’s ability to develop both theoretical frameworks and practical algorithms has shaped modern machine learning, making his work essential reading for students and researchers alike.
Research Focus
Key Achievements
Top Papers
- 1VideoFlow: A Flow-Based Generative Model for Video87 citations · 2019