Re: Relative Entropy
From: illywhacker (illywhacker_at_free.fr)
Date: 02/26/05
- Next message: Josh: "Re: Relative Entropy"
- Previous message: Predictor: "Adjusting"
- In reply to: Josh: "Re: Relative Entropy"
- Next in thread: Josh: "Re: Relative Entropy"
- Reply: Josh: "Re: Relative Entropy"
- Messages sorted by: [ date ] [ thread ]
Date: 26 Feb 2005 05:05:05 -0800
This certainly depends on how you take the limit, i.e. the relationship
between the values of p and q as you take the limit.
However, again at least for finite sets, there is no difference in
practice between a distribution p on some set X and one on the set Z =
X \cup Y if p(Y) = 0. Of course, if q is defined on Z, and q(Y) \neq 0,
you need to include Y. In general, you can restrict p and q to the set
of elements on which one of them in non-zero.
I think the general answer is that you can effectively exclude elements
of the sigma algebra that have measure zero.
illywhacker;
- Next message: Josh: "Re: Relative Entropy"
- Previous message: Predictor: "Adjusting"
- In reply to: Josh: "Re: Relative Entropy"
- Next in thread: Josh: "Re: Relative Entropy"
- Reply: Josh: "Re: Relative Entropy"
- Messages sorted by: [ date ] [ thread ]