Skip to content
ml-recall

#ddim

2 questions from Diffusion Models

Timed test

2 questions · 2 min · score and explanations at the end

Take a test
Q 1 / 2Score 0 / 0

Question 1: Short answer

mediumAI-checkedDiffusion Models

A trained DDPM noise predictor sees the noisy value xt=1.0x_t = 1.0 at a step with αˉt=0.64\bar\alpha_t = 0.64 and predicts ϵθ(xt,t)=1.5\epsilon_\theta(x_t, t) = 1.5. Using xt=αˉt x0+1−αˉt ϵx_t = \sqrt{\bar\alpha_t}\,x_0 + \sqrt{1-\bar\alpha_t}\,\epsilon, what is the model's estimate x^0\hat{x}_0 of the clean value?

a number, 3 decimals