EasyDeepLearn

Score matching — what does it estimate?

hard

Answer

  • Rather than p(x), estimate the score s(x) = ∇_x log p(x).
  • Loss (Hyvärinen): E[sθ(x)      log  p(x)2]E[ \mid \mid s{\theta}(x)\; - \; \nabla \;\operatorname{log}\;p(x) \mid \mid ^{2}] → doesn't need normalization constant → tractable for unnormalized density models.
  • Foundation of energy-based models and diffusion models (score-based generative modeling, Song & Ermon; Ho et al. DDPM).
  • Modern generative modeling paradigm shift.
Check yourself — multiple choice
  • Random
  • Estimate score s(x) = ∇log p(x) — no normalization needed → tractable for unnormalized densities; foundation of energy-based + diffusion models
  • Same as MLE
  • Not real

Score matching: fit ∇log p(x); foundation of diffusion / energy-based models.

#density#deep-learning

Practise Unsupervised Learning

214 interview questions in this topic.

Related questions