azula.guidance.pgdm¶

Pseudo-inverse Guided Diffusion Model (PGDM) internals.

References

Pseudoinverse-Guided Diffusion Models for Inverse Problems (Song et al., 2023)

Classes¶

Creates a PGDM sampler.

class azula.guidance.pgdm.PGDMSampler(denoiser, y, A, A_inv, **kwargs)¶[source]

Creates a PGDM sampler.

Parameters:

denoiser (Denoiser) – A denoiser \(q_\phi(X \mid X_t)\).
y (Tensor) – An observation \(y \sim \mathcal{N}(A(x), \Sigma_y)\).
A (Callable[[Tensor], Tensor]) – The forward operator \(x \mapsto A(x)\).
A_inv (Callable[[Tensor], Tensor]) – The pseudo-inverse operator \(y \mapsto A^\dagger(y)\), such that \(A(A^\dagger(A(x))) = A(x)\).
kwargs – Keyword arguments passed to azula.sample.DDIMSampler.