PlanetMath (more info)
 Math for the people, by the people.
Encyclopedia | Requests | Forums | Docs | Wiki | Random | RSS  
Login
create new user
name:
pass:
forget your password?
Main Menu
Owner confidence rating: Very high Entry average rating: Very high
relative entropy (Definition)

Let $ p$ and $ q$ be probability distributions with supports $ \mathcal{X}$ and $ \mathcal{Y}$ respectively, where $ \mathcal{X} \subset \mathcal{Y}$. The relative entropy or Kullback-Leibler distance between two probability distributions $ p$ and $ q$ is defined as

$\displaystyle D(p\vert\vert q) :=\sum_{x \in \mathcal{X}} p(x) \log \frac{p(x)}{q(x)}.$ (1)

While $ D(p\vert\vert q)$ is often called a distance, it is not a true metric because it is not symmetric and does not satisfy the triangle inequality. However, we do have $ D(p\vert\vert q) \ge 0$ with equality iff $ p = q$.

$\displaystyle -D(p\vert\vert q)$ $\displaystyle = -\sum_{x \in \mathcal{X}} p(x) \log \frac{p(x)}{q(x)}$ (2)
  $\displaystyle = \sum_{x \in \mathcal{X}} p(x) \log \frac{q(x)}{p(x)}$ (3)
  $\displaystyle \le \log \left(\sum_{x \in \mathcal{X}} p(x) \frac{q(x)}{p(x)} \right)$ (4)
  $\displaystyle = \log \left(\sum_{x \in \mathcal{X}} q(x) \right)$ (5)
  $\displaystyle \le \log \left(\sum_{x \in \mathcal{Y}} q(x) \right)$ (6)
  $\displaystyle = 0$ (7)

where the first inequality follows from the concavity of $ \log(x)$ and the second from expanding the sum over the support of $ q$ rather than $ p$.

Relative entropy also comes in a continuous version which looks just as one might expect. For continuous distributions $ f$ and $ g$, $ \mathcal{S}$ the support of $ f$, we have

$\displaystyle D(f\vert\vert g) :=\int_{\mathcal{S}} f \log \frac{f}{g}.$ (8)



"relative entropy" is owned by Mathprof. [ full author list (3) | owner history (3) ]
(view preamble)

View style:

See Also: metric, conditional entropy, mutual information, proof of Gaussian maximizes entropy for given covariance

Other names:  Kullback-Leibler distance
Log in to rate this entry.
(view current ratings)

Cross-references: continuous, sum, inequality, iff, equality, triangle inequality, symmetric, metric, distance, supports, distributions
There are 2 references to this entry.

This is version 7 of relative entropy, born on 2002-02-13, modified 2006-09-21.
Object id is 1945, canonical name is RelativeEntropy.
Accessed 11684 times total.

Classification:
AMS MSC60E05 (Probability theory and stochastic processes :: Distribution theory :: Distributions: general theory)
 94A17 (Information and communication, circuits :: Communication, information :: Measures of information, entropy)

Pending Errata and Addenda
None.
[ View all 7 ]
Discussion
Style: Expand: Order:
forum policy

No messages.

Interact
post | correct | update request | add derivation | add example | add (any)