>"the main effect of introducing
stochasticity is
error correction: diffusion model predictions are approximate, and
noise helps to prevent these approximation errors from accumulating across many sampling steps. In the context of optimisation, the
regularising effect of noise in
stochastic gradient descent (SGD) is well-studied..."
[...]
Variance reduction alone does not explain why distillation of diffusion models is so popular, however. Distillation is also a very effective way to reduce the number of sampling steps required."
What a great article!