Describe the forward and reverse processes in diffusion models.

hard

Answer

$Forward (q): gradually add Gaussian noise to the data over T steps until it becomes ~ N(0, I).$
$Fixed, no learning.$
$x_{t}\; = \;\mathrm{sqrt}(\alpha_{\mathrm{bar}}t) \cdot x_{0}\; + \;\mathrm{sqrt}(1\; - \;\alpha_{\mathrm{bar}}t) \cdot \epsilon$
$(p{\theta})$
$E_{t}$
$At sampling, start from Gaussian noise and denoise for T (or fewer, with DDIM) steps.$

Check yourself — multiple choice

Diffusion doesn't use noise
Forward: add noise until Gaussian; reverse: learn to denoise step-by-step; train ε_θ to predict added noise
Diffusion trains an encoder-decoder
Same as GAN

Diffusion: add noise → train to reverse via noise prediction (DDPM).

#diffusion#generative

214 interview questions in this topic.