Re: Relative Entropy
From: Josh (Josh_at_NOMAIL.NOM)
Date: 02/25/05
- Next message: Reef Fish: "Re: partial correlation and partial correlation coefficient"
- Previous message: Ray Koopman: "Re: Weighted sum of squared normals question"
- In reply to: illywhacker: "Re: Relative Entropy"
- Next in thread: Josh: "Re: Relative Entropy"
- Messages sorted by: [ date ] [ thread ]
Date: Fri, 25 Feb 2005 19:02:27 +0100
illywhacker wrote:
> Just separate the log into two terms: p(x) log p(x) and p(x) log q(x).
> It is equivalent to eliminate this point from the set of xs,
> illywhacker;
>
Ok, but if I consider the relative entropy as a function
D(.,.): A\subseteq [0,1]^n x [0,1]^n --> [0,+infty]
then
A={(p,q)\in [0,1]^n x [0,1]^n :
Not \exists i\in{1,...,n} (p_i,q_i)=(0,0) }
where p_i is the i-th componente of vectore p.
Is this right?
> at least
> in the case of finite sets, since it cannot occur.
I think also infinite numerable, is it right?
Thanks, and excuse me for my english.
- Next message: Reef Fish: "Re: partial correlation and partial correlation coefficient"
- Previous message: Ray Koopman: "Re: Weighted sum of squared normals question"
- In reply to: illywhacker: "Re: Relative Entropy"
- Next in thread: Josh: "Re: Relative Entropy"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
|
