top of page

How I like to think about diffusion

Writer's picture: Ethan SmithEthan Smith

It's a bit hard to see in the diagram but in addition to being convolved with a gaussian, these points are also drifting close to zero.

There's two perspectives here actually. There's setting a point to

xt = x0 * alpha + noise * sigma

where sigma and alpha are both numbers between 0 and 1

and then there's

xt = x0 + noise * sigma

but sigma goes towards infinity at the end of the diffusion schedule.


In both cases we can achieve a desired signal to noise ratio but one case involves reducing the image signal while the other keeps it constant and continues raising the noise's variance to overwhelm the signal entirely. I believe these are the Variance Preserving and Variance Exploding perspectives respectively.


I put together this repo, Boneless Flow, to explore it some more. Instead of training a model, with weights, to obtain estimates of the flow towards the manifold of clean data, we can compute the ground truth flow analytically if we have the whole dataset in memory.


A problem with this is that the ground truth score actually won't allow for generating new samples


9 views0 comments

Recent Posts

See All

Kommentarer


bottom of page