View source for Diederik Kingma

== Diederik Kingma ==
A Dutch researcher responsible for two of deep learning's most-used pieces of machinery. With Max Welling he introduced the Variational Autoencoder in 2013, showing how to train a probabilistic generative model with gradient descent via the reparameterisation trick, and with Jimmy Ba he published Adam in 2014, the optimiser that has since been the default for training neural networks. VAEs gave generative modelling a principled probabilistic footing and their latent-space approach reappears inside modern diffusion systems. He has worked at OpenAI, Google Brain and Anthropic. (See also: [[Max Welling]], [[Generative AI]], [[Gradient descent]], [[Loss function]])

[[Category:People]]
[[Category:Artificial Intelligence]]
[[Category:Deep Learning Pioneers]]