Open jattenberg opened 6 years ago
If my understanding is correct, the predict
function outputs the mean of y's posterior distribution. The mean is not the only choice possible and I would suggest specifying the underlying loss function in the docstring (squared-error loss in this example).
see Bayesian point estimation: https://en.wikipedia.org/wiki/Point_estimation#Bayesian_point_estimation
@jattenberg Are you still working on this?
i wanted to be able to sample predictions