moscot.base.problems.OTProblem.prepare

OTProblem.prepare(xy, x, y, a=None, b=None, **kwargs)

Prepare the OT problem.

Depending on which arguments are passed, the problem_kind will be:

• if only xy is non-empty, problem_kind = 'linear'.

• if only x and y are non-empty, problem_kind = 'quadratic'.

• if all xy, x and y are non-empty, problem_kind = 'quadratic'.

Parameters:

• xy (Mapping[str, Any]) – Geometry defining the linear term. If a non-empty dict, from_adata() will be called.

• x (Mapping[str, Any]) – Geometry defining the source quadratic term. If a non-empty dict, from_adata() will be called.

• y (Mapping[str, Any]) – Geometry defining the target quadratic term. If a non-empty dict, from_adata() will be called.

• a (Union[bool, str, ndarray[Any, dtype[float64]], None]) –

  Source marginals. Valid options are:

  • str - key in obs where the source marginals are stored.

  • bool - if True, estimate the marginals from adata_src, otherwise use uniform marginals.

  • ndarray - array of shape [n,] containing the source marginals.

  • None - uniform marginals.

• b (Union[bool, str, ndarray[Any, dtype[float64]], None]) –

  Target marginals. Valid options are:

  • str - key in obs where the target marginals are stored.

  • bool - if True, estimate the marginals from adata_tgt, otherwise use uniform marginals.

  • ndarray - array of shape [m,] containing the target marginals.

  • None - uniform marginals.

• kwargs (Any) – Keyword arguments for estimate_marginals() when a = True or b = True.

Return type:

OTProblem

Returns:

: Self and updates the following fields:

• xy - geometry of shape [n, m] defining the linear term.

• x - geometry of shape [n, n] defining the source quadratic term.

• y - geometry of shape [m, m] defining the target quadratic term.

• a - source marginals of shape [n,].

• b - target marginals of shape [m,].

• solution - set to None.

• stage - set to 'prepared'.

• problem_kind - kind of the OT problem.