ott.neural.methods.flows.dynamics.ConstantNoiseFlow

ott.neural.methods.flows.dynamics.ConstantNoiseFlow#

class ott.neural.methods.flows.dynamics.ConstantNoiseFlow(sigma)[source]#

Flow with straight paths and constant flow noise \(\sigma\).

Parameters:

sigma (float) – Constant noise used for computing time-independent noise schedule.

Methods

compute_mu_t(t, src, tgt)

Compute the mean of the probability path.

compute_sigma_t(t)

Compute noise of the flow at time \(t\).

compute_ut(t, src, tgt)

Evaluate the conditional vector field.

compute_xt(rng, t, src, tgt)

Sample from the probability path.