Indexed text attachment
entropy-sources-and-channels-in-higher-categories.tex
Source for Entropy, Sources, and Channels in Higher Categories
View paper · Download original file
Lines 1321–1420 of 4249. Source text is displayed literally.
1321\end{proof}1322 1323Part (ii) says that a decoder is a map $d\colon Y\to X$ and that among all maps the pointwise maximizer of1324the Bayesian inverse is best. Part (iii) is the quantitative form of a triviality. If a kernel1325$f\colon X\to Y$ has a deterministic left inverse, $d\circ f=\id_X$, then $x$ is a function of $y$ and1326$H(X\st Y)=0$. Fano's inequality says that if $d\circ f$ is within $\varepsilon$ of the identity, in the1327sense that $d(y)\ne x$ with probability at most $\varepsilon$, then the equivocation is at most1328$\hb(\varepsilon)+\varepsilon\log(n-1)$. Approximate retractions lose approximately nothing.1329\Cref{ch:feinstein} is built on this notion.1330 1331\section{Relative entropy as a functor}1332 1333Entropy is a functor on $\FinProb$ (\cref{thm:functor}) and is characterized by that1334(\cref{thm:BFL}). Relative entropy has the same status on a slightly larger category. Let $\FinStat$ have1335finite probability spaces as objects; a morphism $(X,p)\to(Y,q)$ is a pair $(f,s)$ of a1336measure-preserving map $f\colon X\to Y$ and a kernel $s\colon Y\to X$ with $f\circ s=\id_Y$, thought of as a1337hypothesis about how $x$ is distributed given $f(x)$. Composition is $(g,t)\circ(f,s)=(g\circ f,s\circ t)$.1338 1339\begin{theorem}[Baez--Fritz]\label{thm:BF}1340\Sli[\cite{BF14}] The assignment $(f,s)\mapsto\KL p{s\circ q}$ is a functor $\FinStat\to[0,\infty]$. It is lower1341semicontinuous and convex-linear, and it vanishes when $s$ is the Bayesian inverse of $f$. Every functor1342with these properties is a constant multiple of it.1343\end{theorem}1344 1345Functoriality here is \cref{prop:chainD}. The quantity $\KL p{s\circ q}$ is the information gained on learning1346that the true distribution is $p$, when one's prior was built from the correct marginal $q$ and the1347hypothesis $s$.1348 1349\section{Beyond the elementary inequalities}1350 1351Every inequality proved so far is a consequence of \cref{prop:submodular}, that is, of monotonicity and1352submodularity. Such consequences are called Shannon-type. They are all there is for up to three variables.1353 1354\begin{theorem}\label{thm:nonshannon}1355\Scl[\cite{ZY98,Mat07}] For four or more random variables there are linear inequalities among their joint1356entropies that hold for all distributions and are not consequences of monotonicity and submodularity;1357there are infinitely many independent ones, and the closure of the set of entropy vectors is a convex1358cone that is not polyhedral.1359\end{theorem}1360 1361Khinchin's elementary inequalities are thus the first layer of a structure that is not finitely describable.1362Neither the cohomological account of \cref{ch:cocycle} nor the enriched one of this chapter explains the1363further layers. We record this as a limit of what is understood, not of what was needed: the coding1364theorems use only the first layer.1365 1366\section*{Notes}1367 1368Relative entropy is due to Kullback and Leibler; for the classical theory see \cite{CT06,CK11}. The1369enriched-category view of data processing and \cref{prop:copy} are due to Perrone \cite{Per24}. The1370generalized Shannon inequalities and Feinstein's lemmas are \cite[Ch.~I]{Khi56}, after \cite{Fei54}; Fano's1371inequality is in \cite{Fan61}. Submodularity of entropy and the polymatroid point of view go back to1372Fujishige \cite{Fuj78}; non-Shannon inequalities were discovered by Zhang and Yeung \cite{ZY98}, and the1373non-polyhedrality is due to Mat\'u\v{s} \cite{Mat07}. See \cite{Yeu08} for the whole subject. The category1374$\FinStat$ and \cref{thm:BF} are from \cite{BF14}.1375 1376\section*{Exercises}1377 1378\begin{exercise}1379For $p=(\tfrac12,\tfrac12)$ and $q=(\tfrac14,\tfrac34)$ compute $\KL pq$ and $\KL qp$. Let $f$ be the binary1380symmetric channel that flips its input with probability $0.1$. Compute $\KL{f\circ p}{f\circ q}$ and compare.1381\end{exercise}1382 1383\begin{exercise}1384Prove \cref{prop:chainD}, and show that equality holds in \cref{thm:DPI} iff the two Bayesian inverses agree:1385$r(\,\cdot\st y)=r'(\,\cdot\st y)$ for all $y$ in the support of $f\circ p$.1386\end{exercise}1387 1388\begin{exercise}1389Compute $\TV(\cpy\circ p,\,p\otimes p)$ and $\KL{\cpy\circ p}{p\otimes p}$ for $p=(\tfrac12,\tfrac14,\tfrac14)$.1390For which states on $n$ points is each largest?1391\end{exercise}1392 1393\begin{exercise}\label{exer:meet}1394Let $X,Y$ be bits with joint distribution $r(0,0)=r(1,1)=0.4$, $r(0,1)=r(1,0)=0.1$. Compute $I(X;Y)$ and show1395that $X\wedge Y$ is trivial. Then give an example in which $I(X;Y)=H[X\wedge Y]>0$.1396\end{exercise}1397 1398\begin{exercise}1399Let $X$ be uniform on $n$ points and let $Y=X$ with probability $1-\varepsilon$, and otherwise uniform on the1400other $n-1$ points, with $\varepsilon<1-\tfrac1n$. Show that equality holds in Fano's inequality.1401\end{exercise}1402%% ---- end ch/04-inequalities.tex ----1403\part{Sources}1404%% ---- begin ch/05-sources.tex ----1405\chapter{Sources}\label{ch:sources}1406 1407A source, for Khinchin, is a probability measure $\mu$ on the space $\AZ$ of doubly infinite sequences of1408letters from a finite alphabet $A$; it is stationary when $\mu$ is invariant under the shift. This chapter1409sets up the same objects in a Markov category. The sequence space is an infinite tensor product, a source is1410a state on it, and stationary sources and the maps between them are the states and morphisms of a category1411of dynamical systems. We then define the entropy rate, compute it for Markov chains as Khinchin does in his1412first paper, and read finite-state sources as coalgebras.1413 1414\section{Sequence spaces}1415 1416Let $\BorelStoch$ be the category of standard Borel spaces and Markov kernels; it is a Markov category1417containing $\FinStoch$, and its deterministic morphisms are the measurable maps (\cref{app:toolkit}). For a1418finite alphabet $A$ the space $\AZ$ carries the product $\sigma$-algebra $\FF_A$, generated by the1419\term{cylinders}, the sets of sequences with prescribed letters at finitely many places. For a finite set1420$F\subset\Z$ let $\pi_F\colon\AZ\to A^F$ be the projection.