Idiwon-orisun Generative Models

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

1. Kini gangan ni 'Dimegilio' ti awọn awoṣe wọnyi kọ?
   • Ilọsiwaju ti iwuwo iṣeeṣe log-pẹlu ọwọ si titẹ sii
   • Iwọn didara kan lati 0 si 100 fun aworan kọọkan
   • Nọmba awọn akoko ikẹkọ ti pari
   • Aami iyasọtọ ti aaye data
2. Kini idi ti awọn awoṣe ti o da lori Dimegilio ṣe ikẹkọ lori data ti bajẹ pẹlu ariwo ni awọn iwọn pupọ?
   • Lati jẹ ki awọn aworan wo iṣẹ ọna
   • Nitori Dimegilio naa nira lati ṣe iṣiro nibiti data mimọ ko fọnka, ati ariwo kun ni awọn agbegbe yẹn
   • Lati dinku iwọn faili ti dataset
   • Nitori ariwo iyara soke ni GPU
3. Ilana iṣapẹẹrẹ wo ni awọn awoṣe ti o da lori Dimegilio tete lo lati ṣe ipilẹṣẹ data?
   • Ilọkalẹ lori awọn piksẹli si aworan ibi-afẹde ti o wa titi
   • A nikan matrix isodipupo
   • Langevin dainamiki - igbesẹ pẹlu Dimegilio lakoko ti o nfi ariwo kun
   • Aileto cropping ati flipping
4. Isopọ pataki wo ni iwe Dimegilio-SDE ti 2021 fi idi mulẹ?
   • Wipe awọn GAN nigbagbogbo dara julọ ju itankale lọ
   • Awọn nẹtiwọki nkankikan ko le ṣe awoṣe awọn aworan
   • Awọn awoṣe Dimegilio yẹn ṣiṣẹ nikan ni awọn iwọn meji
   • Awọn awoṣe itankale ati awọn awoṣe ti o da lori Dimegilio jẹ awọn iwo meji ti ilana SDE ti nlọsiwaju kanna
5. Bawo ni asọtẹlẹ ariwo ṣe ni ibatan si sisọ asọtẹlẹ Dimegilio ni ṣiṣakoṣo ibaamu Dimegilio?
   • The conditional denoising-score target is −ε/σ for standardized Gaussian noise; the model estimates the noisy-data score
   • Noise prediction and score estimation are unrelated
   • Ariwo asọtẹlẹ ko ṣee ṣe
   • Dimegilio nigbagbogbo jẹ odo fun data ariwo
6. Kí ni deterministic 'iṣeeṣe sisan ODE' jeki ti o funfun sitokasitik iṣapẹẹrẹ ko ni awọn iṣọrọ pese?
   • Lilo agbara GPU ti o ga julọ
   • Iṣiro iṣeeṣe deede ati ọna iṣapẹẹrẹ ipinnu ipinnu
   • Aami isamisi dataset laifọwọyi
   • Yiyọ ti awọn nilo fun eyikeyi nkankikan nẹtiwọki
7. In score-based generation, what does the reverse-time SDE do?
   • It converts image labels into captions
   • It adds noise to an already generated output only
   • It transforms a prior noise distribution back toward the data distribution
   • It estimates the class label of each pixel
8. In a score-based MRI reconstruction method, how can a learned image prior help recover an undersampled scan?
   • Use the prior alone and ignore measured k-space data
   • Combine the learned score prior with the measurement constraint during posterior sampling
   • Replace all measurements with random noise
   • Treat the score as a diagnosis of the patient