P 3.3

2 views
Skip to first unread message

SumanV...@gmail.com

unread,
Oct 10, 2007, 7:34:12 PM10/10/07
to CS281A: Statistical Learning Theory (Fall 2007)
Should the predictive probability p(x_{new}|x_{1},....,x_{n}) also be
conditioned on a hyper-parameter p(x_{new}|x_{1},....,x_{n},a) as in
3.2?

Percy Liang

unread,
Oct 10, 2007, 9:11:00 PM10/10/07
to cs281a...@googlegroups.com
Yes, the predictive distribution depends on the hyperparameter (a)
use for the prior over the parameters.

-Percy

Reply all
Reply to author
Forward
0 new messages