Jethro's Braindump

Entropy

tags
Information Theory, Gibbs’ Inequality

Definitions

The Shannon information content of an outcome \(x\), measured in bits, is defined to be:

\begin{equation} h(x) = \log_2 \frac{1}{P(x)} \end{equation}

The entropy of an ensemble \(X\) is defined to be the average Shannon information content of an outcome:

\begin{equation} H(X)\equiv \sum_{x \in \mathcal{A}_X} P(x) \log \frac{1}{P(x)} \end{equation}

Entropy is 0 when the outcome is deterministic, and maximized with value \(\log(|\mathcal{A}_X|)\) when the outcomes are uniformly distributed.

The joint entropy of two ensembles \(X, Y\) is:

\begin{equation} H(X,Y) \equiv \sum_{x,y \in \mathcal{A}_x \mathcal{A}_y} P(x,y) \log \frac{1}{P(x,y)} \end{equation}

Entropy is additive if the ensembles are independent:

\begin{equation} H(X,Y) = H(X) + H(Y) \end{equation}

Entropy is decomposable.

Icon by Laymik from The Noun Project. Website built with ♥ with Org-mode, Hugo, and Netlify.