JTLATestMultinomial#

class scio.scores.classification.jtla.JTLATestMultinomial[source]#

Bases: TypedDict

JTLA’s multinomial test syntax.

Precisely describes possible test values for multinomial test in JTLA.

Parameters:

type (Literal["multinomial"]) – The type of test. Must be "multinomial".
mode (MultinomialTestModeLike) – See MultinomialTestMode.
k (int) – Number of neighbors to consider.
prior (Tensor | float, optional) – If not provided, treated as 1.0. Must be broadcastable to (n_classes, n_classes), where each row defines the Dirichlet prior for the corresponding class-conditional test. Ignored in MLE mode.
special_prior (float, optional) – If provided, refers to the epsilon from the “special prior” in the authors’ implementation. In this case, prior is ignored.