If I train a model with Minibatch(es), how do I replace its/their value after approximation so that I can sample the posterior for a fixed and known subset of the data? I think I’m missing something obvious, but am going circles to no avail. Is there an example somewhere that would cover this?

Posterior sample from an approximation with Minibatches?

junpenglao February 3, 2018, 5:02pm 5

Although according to the docstring .set_value() really should work, @ferrine?

Topic		Replies	Views
Updating prior with posterior while using minibatches Questions	0	670	October 15, 2018
Evaluating minibatch trained model on test set Questions	1	1555	March 16, 2018
Sampling after minibatch training Questions	1	1049	November 28, 2017
Minibatch and NUTS Questions	11	3350	June 22, 2018
Dealing with Unbalanced Data with Minibatches Questions	16	3590	February 4, 2020