Re:The physics principle of diffusion inspired modern AI art
by PartiallyTyped on Thursday January 05, 2023 @05:38PM
(#34267934)
As far as I could tell from a skim, they refer to score models, where the score is the density of the input.
We then train a network to estimate the gradient of the score, and add that to the input iteratively.
The gradient of the score always points towards where the score (aka density) will increase, and you essentially apply gradients to walk towards a high density region.
This idea is actually very similar to using gradients to crease poisoned inputs that are falsely predicted.
We then train a network to estimate the gradient of the score, and add that to the input iteratively.
The gradient of the score always points towards where the score (aka density) will increase, and you essentially apply gradients to walk towards a high density region.
This idea is actually very similar to using gradients to crease poisoned inputs that are falsely predicted.