Empirical Gaussian prior built from a prior outcome tensor R0 of shape (L, M, D) or (L, M). Prior means are the centered empirical logits.
R0
(L, M, D)
(L, M)
Readonly
Snake-case alias matching Python's public instance attribute.
Evaluate the prior penalty P(theta) (added to a negative log-likelihood).
P(theta)
Empirical Gaussian prior built from a prior outcome tensor
R0of shape(L, M, D)or(L, M). Prior means are the centered empirical logits.