DDPM: Adding Noise in One Step
Intermediate
The problem, from first principles
A diffusion model learns to reverse a gradual noising process. The paper's key identity lets you jump from the clean image to any noise level in one step, without simulating each intermediate step.
From theory to code
Implement forward_noise(x0, ab, eps), the closed-form noised sample at a given noise level.
Constraints
abis the cumulative signal fraction, between 0 and 1.
Hints
Hint 1
Scale the clean sample by the square root of ab and the noise by the square root of 1 - ab, then add.
Keywords
research-paperscomputer-visiondiffusiongenerative-models
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.