These are chat archives for bayespy/bayespy

23rd
May 2016
Deebul Nair
@deebuls
May 23 2016 13:52

Hi @jluttine . Can we do sequential learning in Bayespy.

 p = Beta([1e-3, 1e-3])
 z = Bernoulli(p, plates=(10,))
 z.observe([0, 1, 1, 1, 0, 1, 1, 1, 0, 1])
 p.update()

so we dont mention the plates . and do

z.observer([0]) ; p.update(); z.observe([1]); p.update ...
Jaakko Luttinen
@jluttine
May 23 2016 14:03
hmm.. one way of doing it is to create a new node for each observation..
but that's not very good if you have large datasets
i'm thinking whether the SVI machinery could be used for that purpose
you could open a feature request issue for this
@deebuls ^