quiz bu lëkkale ak tegtal · Diggante Niveau

Laajteb autoencoder yu wuute

Xoolaat séddaleb VAE, parametrisasioŋ bu bees, yamale KL, tàggat ELBO, jël misaal, ak jëfandikoo diffusion bu nëbbu. Saytu firnde yi ak yaatuwaayam.

Yooni njiit yu jëm ciAutoencodeur yu wuute
Laaj 1 de 8

Lan mooy encoder bi ci VAE biy génne ci benn duggal buñ jox?

Xoolaatal laaj bu nekk8 laaj

Sooy def quiz bi, tontu yi ak leeral yi dañuy feeñ.

  1. Lan mooy encoder bi ci VAE biy génne ci benn duggal buñ jox?
    • Parametru séddale probabilite, lu ci melni moyenne ak variance
    • Benn poñ bu takku ci espace bu nëbbu
    • Etiketu klaas
    • ab kopi buñ defaraat ci dugal bi
  2. Luy njariñu kaf reparametrisasioŋ bi?
    • Ngir gëna yaatal dekodeer bi
    • Ngir may gradient yi ñu jaar ci jéego jël misaal ci anam wu bari
    • Ngir dindi soxla dekodeer
    • Ngir gëna kompresse barab bu nëbbu bi
  3. In the standard Gaussian VAE formulation, what does the KL term encourage?
    • The decoder to predict future input tokens
    • The encoder to remove all randomness
    • The approximate posterior to stay close to the chosen prior, commonly a standard normal
    • Every latent variable to have the same observed value
  4. Fonction objectif bi VAEs di gëna am solo ci diiru tàggat-yaram lañuy woowe:
    • Plafond entropi croisé
    • Perte adversaire
    • Marge kaare moyen
    • Firnde ci suufu (ELBO)
  5. In a Gaussian VAE, what enables generation of a new sample?
    • Sample a latent from the prior and decode it with the learned decoder
    • Copy a training image directly without a model
    • Run the encoder backward with no latent sampling
    • Choose a random class label without decoding
  6. Ci xeetu diffusion bu nëbbu bu melni diffusion stabil, ban cër la VAE di def?
    • Dafay etikete nataali tàggat yi
    • Dafay kompresse nataal yi ci barab bu nëbbu fu diffusion bi di daw, ba noppi dekodewaat ci pixel yi
    • Dafay wecci enkoderu mbind mi
    • Dafay xayma ni bruit bi di doxee
  7. In variational inference, what does the ELBO provide for the marginal log likelihood log p(x)?
    • An exact value of log p(x) for every model
    • A guarantee that each generated sample is realistic
    • A tractable lower bound that can be maximized instead of directly maximizing log p(x)
    • A posterior distribution that requires no inference model
  8. What distribution does a VAE decoder model given a latent vector?
    • The encoder’s approximate posterior q(z|x)
    • The training set’s class-frequency histogram only
    • The optimizer’s learning-rate schedule
    • The likelihood of an observation conditioned on that latent, p(x|z)