espnet2.enh.diffusion.sdes.OUVPSDE

source

class espnet2.enh.diffusion.sdes.OUVPSDE(beta_min, beta_max, stiffness=1, N=1000, **ignored_kwargs)

Bases: SDE

OUVPSDE class.

!!! SGMSE authors observed instabilities around t=0.2. !!!

Construct an Ornstein-Uhlenbeck Variance Preserving SDE:

dx = -1/2 * beta(t) * stiffness * (y-x) dt + sqrt(beta(t)) * dw

with

beta(t) = beta_min + t(beta_max - beta_min)

Note that the “steady-state mean” y is not provided at construction, but must rather be given as an argument to the methods which require it (e.g., sde or marginal_prob).

• Parameters:
  • beta_min – smallest sigma.
  • beta_max – largest sigma.
  • stiffness – stiffness factor of the drift. 1 by default.
  • N – number of discretization steps

property T

End time of the SDE.

copy()

marginal_prob(x0, t, y)

Parameters to determine the marginal distribution of

the SDE, $p_t(x|args)$.

prior_logp(z)

Compute log-density of the prior distribution.

Useful for computing the log-likelihood via probability flow ODE.

• Parameters:z – latent code
• Returns: log probability density

prior_sampling(shape, y)

Generate one sample from the prior distribution,

$p_T(x|args)$ with shape shape.

sde(x, t, y)