EasyDeepLearn

β-VAE — what does the β hyperparameter do?

hard

Answer

  • Modifies ELBO to Eq[log  p(xz)]    β    KL(q    p)E_{q}[\operatorname{log}\;p(x \mid z)]\; - \;{\beta}\; \cdot \;\operatorname{KL}(q\; \mid \mid \;p). β > 1 pushes KL harder → more independent latent factors → disentangled representations (each latent dim captures one factor: pose, color, size).
  • Cost: worse reconstruction.
  • Trade-off between disentanglement (large β) and fidelity (small β).
  • Foundation of disentangled representation learning; extended by FactorVAE, β-TCVAE.
Check yourself — multiple choice
  • Random
  • ELBO = recon - β·KL; β > 1 → more independent latents (disentanglement) at cost of reconstruction; FactorVAE / β-TCVAE extend
  • Same as VAE
  • Not real

β-VAE: β on KL → disentanglement vs reconstruction tradeoff.

#representation-learning#deep-learning

Practise Unsupervised Learning

214 interview questions in this topic.

Related questions