04 — Representation Learning

Learning compact, meaningful representations of data without explicit supervision — autoencoders, variational autoencoders, and contrastive methods.

Notes

• Representation Learning — autoencoders, VAE, contrastive learning, disentanglement
• Representation Learning — Implementation — shallow autoencoder, VAE with reparameterisation trick and ELBO loss, SimCLR NT-Xent loss in PyTorch

Links

← 03 — Density Estimation → 03 — Probabilistic Models

• MLP and Representation Learning
• Latent Variable Models — VAE connection