Index: The Book of Statistical ProofsGeneral Theorems ▷ Information theory ▷ Differential entropy ▷ Addition upon multiplication

Theorem: Let $X$ be a continuous random variable. Then, the differential entropy of $X$ increases additively upon multiplication with a constant:

\[\label{eq:dent-add} \mathrm{h}(aX) = \mathrm{h}(X) + \log |a| \; .\]

Proof: By definition, the differential entropy of $X$ is

\[\label{eq:X-dent} \mathrm{h}(X) = - \int_{\mathcal{X}} p(x) \log p(x) \, \mathrm{d}x\]

where $p(x) = f_X(x)$ is the probability density function of $X$.

Define the mappings between $X$ and $Y = aX$ as

\[\label{eq:X-Y} Y = g(X) = aX \quad \Leftrightarrow \quad X = g^{-1}(Y) = \frac{Y}{a} \; .\]

If $a > 0$, then $g(X)$ is a strictly increasing function, such that the probability density function of $Y$ is

\[\label{eq:Y-pdf-c1} f_Y(y) = f_X(g^{-1}(y)) \, \frac{\mathrm{d}g^{-1}(y)}{\mathrm{d}y} \overset{\eqref{eq:X-Y}}{=} \frac{1}{a} \, f_X\left(\frac{y}{a}\right) \; ;\]

if $a < 0$, then $g(X)$ is a strictly decreasing function, such that the probability density function of $Y$ is

\[\label{eq:Y-pdf-c2} f_Y(y) = - f_X(g^{-1}(y)) \, \frac{\mathrm{d}g^{-1}(y)}{\mathrm{d}y} \overset{\eqref{eq:X-Y}}{=} -\frac{1}{a} \, f_X\left(\frac{y}{a}\right) \; ;\]

thus, we can write

\[\label{eq:Y-pdf} f_Y(y) = \frac{1}{|a|} \, f_X\left(\frac{y}{a}\right) \; .\]

Writing down the differential entropy for $Y$, we have:

\[\label{eq:Y-dent-s1} \begin{split} \mathrm{h}(Y) &= - \int_{\mathcal{Y}} f_Y(y) \log f_Y(y) \, \mathrm{d}y \\ &\overset{\eqref{eq:Y-pdf}}{=} - \int_{\mathcal{Y}} \frac{1}{|a|} \, f_X\left(\frac{y}{a}\right) \log \left[ \frac{1}{|a|} \, f_X\left(\frac{y}{a}\right) \right] \, \mathrm{d}y \end{split}\]

Substituting $x = y/a$, such that $y = ax$, this yields:

\[\label{eq:Y-dent-s2} \begin{split} \mathrm{h}(Y) &= - \int_{\left\lbrace y/a \,|\, y \in {\mathcal{Y}} \right\rbrace} \frac{1}{|a|} \, f_X\left(\frac{ax}{a}\right) \log \left[ \frac{1}{|a|} \, f_X\left(\frac{ax}{a}\right) \right] \, \mathrm{d}(ax) \\ &= - \int_{\mathcal{X}} f_X(x) \log \left[ \frac{1}{|a|} \, f_X(x) \right] \, \mathrm{d}x \\ &= - \int_{\mathcal{X}} f_X(x) \left[ \log f_X(x) - \log |a| \right] \, \mathrm{d}x \\ &= - \int_{\mathcal{X}} f_X(x) \log f_X(x) \, \mathrm{d}x + \log |a| \int_{\mathcal{X}} f_X(x) \, \mathrm{d}x \\ &\overset{\eqref{eq:X-dent}}{=} \mathrm{h}(X) + \log |a| \; . \end{split}\]
Sources:

Metadata: ID: P200 | shortcut: dent-add | author: JoramSoch | date: 2020-12-02, 16:39.