mushi.loss_functions.dkl

dkl(E, X)

Generalized Kullback-Liebler divergence, a Bregman divergence (ignores constant term)

\[\sum_{i,j} \left(X_{i, j} \log \frac{X_{i, j}}{\mathbb{E}[X_{i, j}]} - X_{i, j} + \mathbb{E}[X_{i, j}] \right)\]

Parameters

• E (ndarray) – expectation \(\mathbb{E}[\mathbf X]\)

• X (ndarray) – data \(\mathbf X\)

Return type

float64