2022-05-09
  • Topic: Autoencoder
  • PCA limitations: its linear, sometimes cannot reveal structures of data
  • Autoencoder usage: reduce dimensionality, reconstruct data, generative model
  • PCA vs Autoencoders
  • Application
    • Sparse autoencoder
    • Denoising autoencoders
    • Autoencoder as manifold learners
  • Topic: VAE
  • Goal: generation with autoencoder
  • Auto-encoding Varialtional Bayes, NIPS, 2013
  • VAE eample:
    • encoder: x -> sigma, mu -> z
    • decoder: z -> bernoulli parameters -> ouptut pixel