DDPM stands for: Denoising Diffusion Probabilistic Model

Can you imagine creating perfect photos from scratch?

Imagine you're a photographer stuck in a rainy day, unable to capture the perfect shot.

Think of a diffusion model as a magical assistant that gradually turns a blurry photo into a clear one, step by step, by learning from many good photos.

Example

Start with a blurry image (like a rainy photo), and the assistant refines it through 100 steps until it's a clear photo (like a sunny day photo).

Remember this

A denoising diffusion probabilistic model helps turn blurry images into clear ones by learning from many good photos.

Text adapted from Wikipedia, licensed under CC BY-SA 4.0.

Related concepts

Stable Diffusion

Stable Diffusion generates images from text descriptions

denoising score matching does: learns to denoise, which equals learning the score

Can you clean up a noisy picture?

Langevin dynamics does: adds noise to gradient descent to sample from a distribution

How can we simulate randomness in physics?

the reverse process learns: p_θ(x_{t-1}|x_t)

Can we trace back the roots of life?

Markov chain Monte Carlo

MCMC samples from complex posterior distributions

Metropolis–Hastings algorithm

Metropolis-Hastings algorithm samples from difficult distributions

Swipe through 100 ML concepts daily

Open Pocket Polymath