method

assignment.SoftAssignment.__init__

def __init__(self, *args = (), epsilon: float = DEFAULT_EPSILON, num_iter: int = DEFAULT_NUM_ITER, **kwargs = {})

Initialize the soft-assignment module.

Parameters

NameTypeDescription
epsilon = DEFAULT_EPSILONfloatEntropy-regularization weight. Smaller values yield sharper, more discrete transport plans at the cost of slower Sinkhorn convergence and reduced numerical stability.
num_iter = DEFAULT_NUM_ITERintNumber of Sinkhorn iterations.
*args = ()Positional arguments passed to :class:`.Assignment`.
**kwargs = {}Keyword arguments passed to :class:`.Assignment`.

Source: unitrack/assignment/_soft.py:43