ott.neural.methods

ott.neural.methods#

Monge Gap#

monge_gap.monge_gap(map_fn, reference_points)

Monge gap regularizer [Uscidda and Cuturi, 2023].

monge_gap.monge_gap_from_samples(source, target)

Monge gap, instantiated in terms of samples before / after applying map.

monge_gap.MongeGapEstimator(dim_data, model)

Mapping estimator between probability measures.

Neural Dual#

neuraldual.W2NeuralDual(dim_data[, ...])

Solver for the Wasserstein-2 Kantorovich dual between Euclidean spaces.

expectile_neural_dual.ExpectileNeuralDual(...)

Expectile-regularized Neural Optimal Transport (ENOT) [Buzun et al., 2024].

expectile_neural_dual.ENOTPotentials(grad_f, ...)

The dual potentials of the ENOT method [Buzun et al., 2024].

ott.neural.methods.flows#

Flows#

otfm.OTFlowMatching(vf, flow[, match_fn, ...])

(Optimal transport) flow matching [Lipman et al., 2022].

genot.GENOT(vf, flow, data_match_fn, *, ...)

Generative Entropic Neural Optimal Transport [Klein et al., 2023].

dynamics.BaseFlow(sigma)

Base class for all flows.

dynamics.StraightFlow(sigma)

Base class for flows with straight paths.

dynamics.ConstantNoiseFlow(sigma)

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

dynamics.BrownianBridge(sigma)

Brownian Bridge.
Contents