R parity for morie.fn.diffu.diffusion_forward.
Usage
morie_diffu_diffusion_forward(
x0,
t,
betas = NULL,
num_steps = 1000L,
noise = NULL,
seed = 0L
)
Arguments
- x0
Clean sample.
- t
Diffusion timestep (1..num_steps).
- betas
Optional custom \(\beta\) schedule.
- num_steps
Total diffusion steps (default 1000).
- noise
Pre-generated Gaussian noise.
- seed
RNG seed.
Value
Named list (x_t, estimate, noise, alpha_bar, beta, method).
Details
$$x_t = \sqrt{\bar\alpha_t}\, x_0 +
\sqrt{1 - \bar\alpha_t}\, \varepsilon$$
with linear \(\beta\) schedule from 1e-4 to 0.02.
References
Ho, Jain & Abbeel (2020), NeurIPS.
Examples
# See the package vignettes for usage examples:
# vignette(package = "rmorie")