Iyatọ Autoencoders adanwo

Awọn idahun ati awọn alaye han bi o ṣe n mu ibeere naa.

1. Kini koodu koodu inu VAE kan fun titẹ sii kan?
   • Awọn paramita ti pinpin iṣeeṣe kan, gẹgẹbi arosọ ati iyatọ
   • Ojuami ti o wa titi kan ni aaye wiwakọ
   • Aami kilasi kan
   • A tunkọ daakọ ti awọn igbewọle
2. Kini idi ti ẹtan reparameterization?
   • Lati jẹ ki oluyipada naa tobi
   • Lati gba awọn gradients laaye lati ṣàn nipasẹ ipele iṣapẹẹrẹ laileto
   • Lati yọ awọn nilo fun a decoder
   • Lati funmorawon aaye wiwaba siwaju
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. Iṣẹ ibi-afẹde ti VAEs pọ si lakoko ikẹkọ ni a pe ni:
   • Cross-entropy aja
   • adanu ọta
   • Itumọ ala onigun mẹrin
   • Ẹri Isalẹ Ilẹ (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. Ninu awoṣe itọka wiwaba bi Stable Diffusion, ipa wo ni VAE ṣe?
   • O ṣe aami awọn aworan ikẹkọ
   • O fun awọn aworan pọ si aaye wiwaba nibiti itankale n ṣiṣẹ, lẹhinna pinnu pada si awọn piksẹli
   • O rọpo koodu koodu
   • O ṣeto awọn ipele ariwo
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)