Variational autoencoders. Learn with Zon & Pavan.

Start with one tiny picture and follow it through an encoder, a Gaussian code distribution and a decoder.

Twelve spoken conversations with Zon and Pavan encourage the learner to predict, pause, explore and explain. Readable transcripts and the longer supplied narration accompany each step.

Explore reconstruction, sampling, KL divergence, the latent map and a comparison with a plain autoencoder. These controls inspect pretrained models rather than train new weights.

The complete lesson and its audio use one VAE access permission. Sign in with approved access or request access.

Explore next

• Explore Transformers
• Backpropagation with real numbers
• Logistic regression foundations

Stories, illustrations, and interactive mini lessons. Ask a question.