moscot.problems.generic.SinkhornProblem#

class moscot.problems.generic.SinkhornProblem(adata, **kwargs)[source]#

Class for solving a linear problem.

Parameters:

Methods

add_problem(key, problem, *[, overwrite])

Add a subproblem.

cell_transition(source, target[, ...])

Aggregate the transport matrix.

compute_entropy(source, target[, forward, ...])

Compute the conditional entropy per cell.

compute_feature_correlation(obs_key[, ...])

Compute correlation of push-forward or pull-back distribution with features.

load(path)

Load the model from a file.

prepare(key[, joint_attr, policy, cost, ...])

Prepare the individual linear subproblems.

pull(source, target[, data, subset, ...])

Pull mass from target to source.

push(source, target[, data, subset, ...])

Push mass from source to target.

remove_problem(key)

Remove a subproblem.

save(path[, overwrite])

Save the problem to a file.

solve([epsilon, tau_a, tau_b, rank, ...])

Solve the individual linear subproblems using the Sinkhorn algorithm [Cuturi, 2013].

Attributes

adata

Annotated data object.

batch_key

Batch key in obs.

problem_kind

Kind of the underlying problem.

problems

OT subproblems that define the biological problem.

solutions

Solutions to the problems.

stage

Problem stage.