moscot.problems.spatiotemporal.SpatioTemporalProblem.compute_random_distance

SpatioTemporalProblem.compute_random_distance(source, intermediate, target, interpolation_parameter=None, n_interpolated_cells=None, account_for_unbalancedness=False, posterior_marginals=True, seed=None, backend='ott', **kwargs)

Compute Wasserstein distance between randomly interpolated and intermediate cells.

See also

  • TODO(MUCDK): create an example showing the usage.

This function interpolates cells between the source and target distributions using a random OT coupling to approximate cells at the intermediate time point.

Parameters:

  • source (TypeVar(K, bound= Hashable)) – Key identifying the source distribution.

  • intermediate (TypeVar(K, bound= Hashable)) – Key identifying the intermediate distribution.

  • target (TypeVar(K, bound= Hashable)) – Key identifying the target distribution.

  • interpolation_parameter (Optional[float]) – Interpolation parameter in \((0, 1)\) defining the weight of the source and target distributions. If None, it is linearly interpolated.

  • n_interpolated_cells (Optional[int]) – Number of cells used for interpolation. If None, use the number of cells in the intermediate distribution.

  • account_for_unbalancedness (bool) – Whether to account for unbalancedness by assuming exponential cell growth and death.

  • posterior_marginals (bool) – Whether to use posterior_growth_rates or prior_growth_rates. TODO(MUCDK): needs more explanation

  • seed (Optional[int]) – Random seed used when sampling the interpolated cells.

  • backend (Literal['ott']) – Backend used for the distance computation.

  • **kwargs (Any) – The description is missing.

Return type:

float

Returns:

: The distance between randomly interpolated cells and cells at the intermediate time point.