Artificial intelligence · 2013
Variational autoencoder
A network that learns a smooth latent space it can sample new data from.
Diederik Kingma and Max Welling's VAE combined neural networks with variational inference via the reparameterisation trick. VAEs made deep generative modelling practical, and their encoders later compressed images for Stable Diffusion's latent space.
What it built on 2
What it led to 4
Related measurements
The dashed line marks when it first worked.
Training compute of notable AI models Source: Epoch AI, Notable AI Models (CC BY 4.0)
ML hardware FLOP/s per dollar Source: Epoch AI, ML hardware (CC BY 4.0)