Index: The Book of Statistical ProofsGeneral Theorems ▷ Information theory ▷ Kullback-Leibler divergence ▷ Relation to differential entropy

Theorem: Let $X$ be a continuous random variable with possible outcomes $\mathcal{X}$ and let $P$ and $Q$ be two probability distributions on $X$. Then, the Kullback-Leibler divergence of $P$ from $Q$ can be expressed as

\[\label{eq:kl-dent} \mathrm{KL}[P||Q] = \mathrm{h}(P,Q) - \mathrm{h}(P)\]

where $\mathrm{h}(P,Q)$ is the differential cross-entropy of $P$ and $Q$ and $\mathrm{h}(P)$ is the marginal differential entropy of $P$.

Proof: The continuous Kullback-Leibler divergence is defined as

\[\label{eq:KL} \mathrm{KL}[P||Q] = \int_{\mathcal{X}} p(x) \cdot \log \frac{p(x)}{q(x)} \, \mathrm{d}x\]

where $p(x)$ and $q(x)$ are the probability density functions of $P$ and $Q$.

Separating the logarithm, we have:

\[\label{eq:KL-dev} \mathrm{KL}[P||Q] = - \int_{\mathcal{X}} p(x) \, \log q(x) \, \mathrm{d}x + \int_{\mathcal{X}} p(x) \, \log p(x) \, \mathrm{d}x \; .\]

Now considering the definitions of marginal differential entropy and differential cross-entropy

\[\label{eq:MDE-DCE} \begin{split} \mathrm{h}(P) &= - \int_{\mathcal{X}} p(x) \, \log p(x) \, \mathrm{d}x \\ \mathrm{h}(P,Q) &= - \int_{\mathcal{X}} p(x) \, \log q(x) \, \mathrm{d}x \; , \end{split}\]

we can finally show:

\[\label{eq:KL-qed} \mathrm{KL}[P||Q] = \mathrm{h}(P,Q) - \mathrm{h}(P) \; .\]
Sources:

Metadata: ID: P114 | shortcut: kl-dent | author: JoramSoch | date: 2020-05-27, 23:32.