>"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."
>"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!