Relation of continuous mutual information to marginal and joint differential entropy

Index: The Book of Statistical Proofs ▷ General Theorems ▷ Information theory ▷ Continuous mutual information ▷ Relation to marginal and joint differential entropy

Theorem: Let $X$ and $Y$ be continuous random variables with the joint probability $p(x,y)$ for $x \in \mathcal{X}$ and $y \in \mathcal{Y}$. Then, the mutual information of $X$ and $Y$ can be expressed as

\[\label{eq:cmi-mjde} \mathrm{I}(X,Y) = \mathrm{h}(X) + \mathrm{h}(Y) - \mathrm{h}(X,Y)\]

where $\mathrm{h}(X)$ and $\mathrm{h}(Y)$ are the marginal differential entropies of $X$ and $Y$ and $\mathrm{h}(X,Y)$ is the joint differential entropy.

Proof: The mutual information of $X$ and $Y$ is defined as

\[\label{eq:MI} \mathrm{I}(X,Y) = \int_{\mathcal{X}} \int_{\mathcal{Y}} p(x,y) \log \frac{p(x,y)}{p(x)\,p(y)} \, \mathrm{d}y \, \mathrm{d}x \; .\]

Separating the logarithm, we have:

\[\label{eq:MI-s1} \mathrm{I}(X,Y) = \int_{\mathcal{X}} \int_{\mathcal{Y}} p(x,y) \log p(x,y) \, \mathrm{d}y \, \mathrm{d}x - \int_{\mathcal{X}} \int_{\mathcal{Y}} p(x,y) \log p(x) \, \mathrm{d}y \, \mathrm{d}x - \int_{\mathcal{X}} \int_{\mathcal{Y}} p(x,y) \log p(y) \, \mathrm{d}y \, \mathrm{d}x \; .\]

Regrouping the variables, this reads:

\[\label{eq:MI-s2} \mathrm{I}(X,Y) = \int_{\mathcal{X}} \int_{\mathcal{Y}} p(x,y) \log p(x,y) \, \mathrm{d}y \, \mathrm{d}x - \int_{\mathcal{X}} \left( \int_{\mathcal{Y}} p(x,y) \, \mathrm{d}y \right) \log p(x) \, \mathrm{d}x - \int_{\mathcal{Y}} \left( \int_{\mathcal{X}} p(x,y) \, \mathrm{d}x \right) \log p(y) \, \mathrm{d}y \; .\]

Applying the law of marginal probability, i.e. $p(x) = \int_{\mathcal{Y}} p(x,y)$, we get:

\[\label{eq:MI-s3} \mathrm{I}(X,Y) = \int_{\mathcal{X}} \int_{\mathcal{Y}} p(x,y) \log p(x,y) \, \mathrm{d}y \, \mathrm{d}x - \int_{\mathcal{X}} p(x) \log p(x) \, \mathrm{d}x - \int_{\mathcal{Y}} p(y) \log p(y) \, \mathrm{d}y \; .\]

Now considering the definitions of marginal and joint differential entropy

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

we can finally show:

\[\label{eq:MI-qed} \begin{split} \mathrm{I}(X,Y) &= - \mathrm{h}(X,Y) + \mathrm{h}(X) + \mathrm{h}(Y) \\ &= \mathrm{h}(X) + \mathrm{h}(Y) - \mathrm{h}(X,Y) \; . \end{split}\]

∎

Sources:

Wikipedia (2020): "Mutual information"; in: Wikipedia, the free encyclopedia, retrieved on 2020-02-21; URL: https://en.wikipedia.org/wiki/Mutual_information#Relation_to_conditional_and_joint_entropy.

Metadata: ID: P59 | shortcut: cmi-mjde | author: JoramSoch | date: 2020-02-21, 17:13.