Movatterモバイル変換

[0]ホーム
URL:

Jump to content
WikipediaThe Free Encyclopedia
Search

Entropy (information theory)

From Wikipedia, the free encyclopedia
Expected amount of information needed to specify the output of a stochastic data source
For other uses, seeEntropy (disambiguation).
icon
This articleneeds additional citations forverification. Please helpimprove this article byadding citations to reliable sources. Unsourced material may be challenged and removed.
Find sources: "Entropy" information theory – news ·newspapers ·books ·scholar ·JSTOR(February 2019) (Learn how and when to remove this message)

Information theory

Ininformation theory, theentropy of arandom variable quantifies the average level of uncertainty or information associated with the variable's potential states or possible outcomes. This measures the expected amount of information needed to describe the state of the variable, considering the distribution of probabilities across all potential states. Given a discrete random variableX{\displaystyle X}, which may be any memberx{\displaystyle x} within the setX{\displaystyle {\mathcal {X}}} and is distributed according top:X→[0,1]{\displaystyle p\colon {\mathcal {X}}\to [0,1]}, the entropy isH(X):=−∑x∈Xp(x)log⁡p(x),{\displaystyle \mathrm {H} (X):=-\sum _{x\in {\mathcal {X}}}p(x)\log p(x),}whereΣ{\displaystyle \Sigma } denotes the sum over the variable's possible values.[Note 1] The choice of base forlog{\displaystyle \log }, thelogarithm, varies for different applications. Base 2 gives the unit ofbits (or "shannons"), while basee gives "natural units"nat, and base 10 gives units of "dits", "bans", or "hartleys". An equivalent definition of entropy is theexpected value of theself-information of a variable.[1]

Two bits of entropy: In the case of two fair coin tosses, the information entropy in bits is the base-2 logarithm of the number of possible outcomes‍— with two coins there are four possible outcomes, and two bits of entropy. Generally, information entropy is the average amount of information conveyed by an event, when considering all possible outcomes.

The concept of information entropy was introduced byClaude Shannon in his 1948 paper "A Mathematical Theory of Communication",[2][3] and is also referred to asShannon entropy. Shannon's theory defines adata communication system composed of three elements: a source of data, acommunication channel, and a receiver. The "fundamental problem of communication" – as expressed by Shannon – is for the receiver to be able to identify what data was generated by the source, based on the signal it receives through the channel.[2][3] Shannon considered various ways to encode, compress, and transmit messages from a data source, and proved in hissource coding theorem that the entropy represents an absolute mathematical limit on how well data from the source can belosslessly compressed onto a perfectly noiseless channel. Shannon strengthened this result considerably for noisy channels in hisnoisy-channel coding theorem.

Entropy in information theory is directly analogous to theentropy instatistical thermodynamics. The analogy results when the values of the random variable designate energies of microstates, so Gibbs's formula for the entropy is formally identical to Shannon's formula. Entropy has relevance to other areas of mathematics such ascombinatorics andmachine learning. The definition can be derived from a set ofaxioms establishing that entropy should be a measure of how informative the average outcome of a variable is. For a continuous random variable,differential entropy is analogous to entropy. The definitionE[−log⁡p(X)]{\displaystyle \mathbb {E} [-\log p(X)]} generalizes the above.

Introduction

[edit]

The core idea of information theory is that the "informational value" of a communicated message depends on the degree to which the content of the message is surprising. If a highly likely event occurs, the message carries very little information. On the other hand, if a highly unlikely event occurs, the message is much more informative. For instance, the knowledge that some particular numberwill not be the winning number of a lottery provides very little information, because any particular chosen number will almost certainly not win. However, knowledge that a particular numberwill win a lottery has high informational value because it communicates the occurrence of a very low probability event.

Theinformation content, also called thesurprisal orself-information, of an eventE{\displaystyle E} is a function that increases as the probabilityp(E){\displaystyle p(E)} of an event decreases. Whenp(E){\displaystyle p(E)} is close to 1, the surprisal of the event is low, but ifp(E){\displaystyle p(E)} is close to 0, the surprisal of the event is high. This relationship is described by the functionlog⁡(1p(E)),{\displaystyle \log \left({\frac {1}{p(E)}}\right),}wherelog{\displaystyle \log } is thelogarithm, which gives 0 surprise when the probability of the event is 1.[4] In fact,log is the only function that satisfies а specific set of conditions defined in section§ Characterization.

Hence, we can define the information, or surprisal, of an eventE{\displaystyle E} by

I(E)=log⁡(1p(E)),{\displaystyle I(E)=\log \left({\frac {1}{p(E)}}\right),}or equivalently,I(E)=−log⁡(p(E)).{\displaystyle I(E)=-\log(p(E)).}

Entropy measures the expected (i.e., average) amount of information conveyed by identifying the outcome of a random trial.[5]: 67  This implies that rolling a dice has higher entropy than tossing a coin because each outcome of a dice toss has smaller probability (p=1/6{\displaystyle p=1/6}) than each outcome of a coin toss (p=1/2{\displaystyle p=1/2}).

Consider a coin with probabilityp of landing on heads and probability1 −p of landing on tails. The maximum surprise is whenp = 1/2, for which one outcome is not expected over the other. In this case a coin flip has an entropy of onebit (similarly, onetrit with equiprobable values containslog2⁡3{\displaystyle \log _{2}3} (about 1.58496) bits of information because it can have one of three values). The minimum surprise is whenp = 0 (impossibility) orp = 1 (certainty) and the entropy is zero bits. When the entropy is zero, there is no uncertainty at all – no freedom of choice – noinformation.[6] Other values ofp give entropies between zero and one bits.

Example

[edit]

Information theory is useful to calculate the smallest amount of information required to convey a message, as indata compression. For example, consider the transmission of sequences comprising the 4 characters 'A', 'B', 'C', and 'D' over a binary channel. If all 4 letters are equally likely (25%), one cannot do better than using two bits to encode each letter. 'A' might code as '00', 'B' as '01', 'C' as '10', and 'D' as '11'. However, if the probabilities of each letter are unequal, say 'A' occurs with 70% probability, 'B' with 26%, and 'C' and 'D' with 2% each, one could assign variable length codes. In this case, 'A' would be coded as '0', 'B' as '10', 'C' as '110', and 'D' as '111'. With this representation, 70% of the time only one bit needs to be sent, 26% of the time two bits, and only 4% of the time 3 bits. On average, fewer than 2 bits are required since the entropy is lower (owing to the high prevalence of 'A' followed by 'B' – together 96% of characters). The calculation of the sum of probability-weighted log probabilities measures and captures this effect.

English text, treated as a string of characters, has fairly low entropy; i.e. it is fairly predictable. We can be fairly certain that, for example, 'e' will be far more common than 'z', that the combination 'qu' will be much more common than any other combination with a 'q' in it, and that the combination 'th' will be more common than 'z', 'q', or 'qu'. After the first few letters one can often guess the rest of the word. English text has between 0.6 and 1.3 bits of entropy per character of the message.[7]: 234 

Definition

[edit]

Named afterBoltzmann's Η-theorem, Shannon defined the entropyΗ (Greek capital lettereta) of adiscrete random variableX{\textstyle X}, which takes values in the setX{\displaystyle {\mathcal {X}}} and is distributed according top:X→[0,1]{\displaystyle p:{\mathcal {X}}\to [0,1]} such thatp(x):=P[X=x]{\displaystyle p(x):=\mathbb {P} [X=x]}:

H(X)=E[I⁡(X)]=E[−log⁡p(X)].{\displaystyle \mathrm {H} (X)=\mathbb {E} [\operatorname {I} (X)]=\mathbb {E} [-\log p(X)].}

HereE{\displaystyle \mathbb {E} } is theexpected value operator, andI is theinformation content ofX.[8]: 11 [9]: 19–20 I⁡(X){\displaystyle \operatorname {I} (X)} is itself a random variable.

The entropy can explicitly be written as:H(X)=−∑x∈Xp(x)logb⁡p(x),{\displaystyle \mathrm {H} (X)=-\sum _{x\in {\mathcal {X}}}p(x)\log _{b}p(x),}whereb is thebase of the logarithm used. Common values ofb are 2,Euler's numbere, and 10, and the corresponding units of entropy are thebits forb = 2,nats forb =e, andbans forb = 10.

In the case ofp(x)=0{\displaystyle p(x)=0} for somex∈X{\displaystyle x\in {\mathcal {X}}}, the value of the corresponding summand0 logb(0) is taken to be0, which is consistent with thelimit:[10]: 13 limp→0+plog⁡(p)=0.{\displaystyle \lim _{p\to 0^{+}}p\log(p)=0.}

One may also define theconditional entropy of two variablesX{\displaystyle X} andY{\displaystyle Y} taking values from setsX{\displaystyle {\mathcal {X}}} andY{\displaystyle {\mathcal {Y}}} respectively, as:[10]: 16 H(X|Y)=−∑x,y∈X×YpX,Y(x,y)log⁡pX,Y(x,y)pY(y),{\displaystyle \mathrm {H} (X|Y)=-\sum _{x,y\in {\mathcal {X}}\times {\mathcal {Y}}}p_{X,Y}(x,y)\log {\frac {p_{X,Y}(x,y)}{p_{Y}(y)}},}wherepX,Y(x,y):=P[X=x,Y=y]{\displaystyle p_{X,Y}(x,y):=\mathbb {P} [X=x,Y=y]} andpY(y)=P[Y=y]{\displaystyle p_{Y}(y)=\mathbb {P} [Y=y]}. This quantity should be understood as the remaining randomness in the random variableX{\displaystyle X} given the random variableY{\displaystyle Y}.

Measure theory

[edit]

Entropy can be formally defined in the language ofmeasure theory as follows:[11] Let(X,Σ,μ){\displaystyle (X,\Sigma ,\mu )} be aprobability space. LetA∈Σ{\displaystyle A\in \Sigma } be anevent. Thesurprisal ofA{\displaystyle A} isσμ(A)=−ln⁡μ(A).{\displaystyle \sigma _{\mu }(A)=-\ln \mu (A).}

Theexpected surprisal ofA{\displaystyle A} ishμ(A)=μ(A)σμ(A).{\displaystyle h_{\mu }(A)=\mu (A)\sigma _{\mu }(A).}

Aμ{\displaystyle \mu }-almostpartition is aset familyP⊆P(X){\displaystyle P\subseteq {\mathcal {P}}(X)} such thatμ(∪⁡P)=1{\displaystyle \mu (\mathop {\cup } P)=1} andμ(A∩B)=0{\displaystyle \mu (A\cap B)=0} for all distinctA,B∈P{\displaystyle A,B\in P}. (This is a relaxation of the usual conditions for a partition.) The entropy ofP{\displaystyle P} isHμ(P)=∑A∈Phμ(A).{\displaystyle \mathrm {H} _{\mu }(P)=\sum _{A\in P}h_{\mu }(A).}

LetM{\displaystyle M} be asigma-algebra onX{\displaystyle X}. The entropy ofM{\displaystyle M} isHμ(M)=supP⊆MHμ(P).{\displaystyle \mathrm {H} _{\mu }(M)=\sup _{P\subseteq M}\mathrm {H} _{\mu }(P).}Finally, the entropy of the probability space isHμ(Σ){\displaystyle \mathrm {H} _{\mu }(\Sigma )}, that is, the entropy with respect toμ{\displaystyle \mu } of the sigma-algebra ofall measurable subsets ofX{\displaystyle X}.

Example

[edit]
EntropyΗ(X) (i.e. theexpectedsurprisal) of a coin flip, measured in bits, graphed versus the bias of the coinPr(X = 1), whereX = 1 represents a result of heads.[10]: 14–15 

Here, the entropy is at most 1 bit, and to communicate the outcome of a coin flip (2 possible values) will require an average of at most 1 bit (exactly 1 bit for a fair coin). The result of a fair die (6 possible values) would have entropy log26 bits.
Main articles:Binary entropy function andBernoulli process

Consider tossing a coin with known, not necessarily fair, probabilities of coming up heads or tails; this can be modeled as aBernoulli process.

The entropy of the unknown result of the next toss of the coin is maximized if the coin is fair (that is, if heads and tails both have equal probability 1/2). This is the situation of maximum uncertainty as it is most difficult to predict the outcome of the next toss; the result of each toss of the coin delivers one full bit of information. This is becauseH(X)=−∑i=1np(xi)logb⁡p(xi)=−∑i=1212log2⁡12=−∑i=1212⋅(−1)=1.{\displaystyle {\begin{aligned}\mathrm {H} (X)&=-\sum _{i=1}^{n}{p(x_{i})\log _{b}p(x_{i})}\\&=-\sum _{i=1}^{2}{{\frac {1}{2}}\log _{2}{\frac {1}{2}}}\\&=-\sum _{i=1}^{2}{{\frac {1}{2}}\cdot (-1)}=1.\end{aligned}}}

However, if we know the coin is not fair, but comes up heads or tails with probabilitiesp andq, wherep ≠q, then there is less uncertainty. Every time it is tossed, one side is more likely to come up than the other. The reduced uncertainty is quantified in a lower entropy: on average each toss of the coin delivers less than one full bit of information. For example, ifp = 0.7, thenH(X)=−plog2⁡p−qlog2⁡q=−0.7log2⁡(0.7)−0.3log2⁡(0.3)≈−0.7⋅(−0.515)−0.3⋅(−1.737)=0.8816<1.{\displaystyle {\begin{aligned}\mathrm {H} (X)&=-p\log _{2}p-q\log _{2}q\\[1ex]&=-0.7\log _{2}(0.7)-0.3\log _{2}(0.3)\\[1ex]&\approx -0.7\cdot (-0.515)-0.3\cdot (-1.737)\\[1ex]&=0.8816<1.\end{aligned}}}

Uniform probability yields maximum uncertainty and therefore maximum entropy. Entropy, then, can only decrease from the value associated with uniform probability. The extreme case is that of a double-headed coin that never comes up tails, or a double-tailed coin that never results in a head. Then there is no uncertainty. The entropy is zero: each toss of the coin delivers no new information as the outcome of each coin toss is always certain.[10]: 14–15 

Characterization

[edit]

To understand the meaning of−Σpi log(pi), first define an information functionI in terms of an eventi with probabilitypi. The amount of information acquired due to the observation of eventi follows from Shannon's solution of the fundamental properties ofinformation:[12]

  1. I(p) ismonotonically decreasing inp: an increase in the probability of an event decreases the information from an observed event, and vice versa.
  2. I(1) = 0: events that always occur do not communicate information.
  3. I(p1·p2) = I(p1) + I(p2): the information learned fromindependent events is the sum of the information learned from each event.
  4. I(p) is a twice continuously differentiable function of p.

Given two independent events, if the first event can yield one ofnequiprobable outcomes and another has one ofm equiprobable outcomes then there aremn equiprobable outcomes of the joint event. This means that iflog2(n) bits are needed to encode the first value andlog2(m) to encode the second, one needslog2(mn) = log2(m) + log2(n) to encode both.

Shannon discovered that a suitable choice ofI{\displaystyle \operatorname {I} } is given by:[13]I⁡(p)=log⁡(1p)=−log⁡(p).{\displaystyle \operatorname {I} (p)=\log \left({\tfrac {1}{p}}\right)=-\log(p).}

In fact, the only possible values ofI{\displaystyle \operatorname {I} } areI⁡(u)=klog⁡u{\displaystyle \operatorname {I} (u)=k\log u} fork<0{\displaystyle k<0}. Additionally, choosing a value fork is equivalent to choosing a valuex>1{\displaystyle x>1} fork=−1/log⁡x{\displaystyle k=-1/\log x}, so thatx corresponds to thebase for the logarithm. Thus, entropy ischaracterized by the above four properties.

Proof
LetI{\textstyle \operatorname {I} } be the information function which one assumes to be twice continuously differentiable, one has:

I⁡(p1p2)= I⁡(p1)+I⁡(p2)Starting from property 3p2I′⁡(p1p2)= I′⁡(p1)taking the derivative w.r.t p1I′⁡(p1p2)+p1p2I″⁡(p1p2)= 0taking the derivative w.r.t p2I′⁡(u)+uI″⁡(u)= 0introducingu=p1p2(uI′⁡(u))′= 0combining terms into one uI′⁡(u)−k= 0integrating w.r.t u,producing constantk{\displaystyle {\begin{aligned}&\operatorname {I} (p_{1}p_{2})&=\ &\operatorname {I} (p_{1})+\operatorname {I} (p_{2})&&\quad {\text{Starting from property 3}}\\&p_{2}\operatorname {I} '(p_{1}p_{2})&=\ &\operatorname {I} '(p_{1})&&\quad {\text{taking the derivative w.r.t}}\ p_{1}\\&\operatorname {I} '(p_{1}p_{2})+p_{1}p_{2}\operatorname {I} ''(p_{1}p_{2})&=\ &0&&\quad {\text{taking the derivative w.r.t}}\ p_{2}\\&\operatorname {I} '(u)+u\operatorname {I} ''(u)&=\ &0&&\quad {\text{introducing}}\,u=p_{1}p_{2}\\&(u\operatorname {I} '(u))'&=\ &0&&\quad {\text{combining terms into one}}\ \\&u\operatorname {I} '(u)-k&=\ &0&&\quad {\text{integrating w.r.t}}\ u,{\text{producing constant}}\,k\\\end{aligned}}}

Thisdifferential equation leads to the solutionI⁡(u)=klog⁡u+c{\displaystyle \operatorname {I} (u)=k\log u+c} for somek,c∈R{\displaystyle k,c\in \mathbb {R} }. Property 2 givesc=0{\displaystyle c=0}. Property 1 and 2 give thatI⁡(p)≥0{\displaystyle \operatorname {I} (p)\geq 0} for allp∈[0,1]{\displaystyle p\in [0,1]}, so thatk<0{\displaystyle k<0}.

The differentunits of information (bits for thebinary logarithmlog2,nats for thenatural logarithmln,bans for thedecimal logarithmlog10 and so on) areconstant multiples of each other. For instance, in case of a fair coin toss, heads provideslog2(2) = 1 bit of information, which is approximately 0.693 nats or 0.301 decimal digits. Because of additivity,n tosses providen bits of information, which is approximately0.693n nats or0.301n decimal digits.

Themeaning of the events observed (the meaning ofmessages) does not matter in the definition of entropy. Entropy only takes into account the probability of observing a specific event, so the information it encapsulates is information about the underlyingprobability distribution, not the meaning of the events themselves.

Alternative characterization

[edit]

Another characterization of entropy uses the following properties. We denotepi = Pr(X =xi) andΗn(p1, ...,pn) = Η(X).

  1. Continuity:H should becontinuous, so that changing the values of the probabilities by a very small amount should only change the entropy by a small amount.
  2. Symmetry:H should be unchanged if the outcomesxi are re-ordered. That is,Hn(p1,p2,…,pn)=Hn(pi1,pi2,…,pin){\displaystyle \mathrm {H} _{n}\left(p_{1},p_{2},\ldots ,p_{n}\right)=\mathrm {H} _{n}\left(p_{i_{1}},p_{i_{2}},\ldots ,p_{i_{n}}\right)} for anypermutation{i1,...,in}{\displaystyle \{i_{1},...,i_{n}\}} of{1,...,n}{\displaystyle \{1,...,n\}}.
  3. Maximum:Hn{\displaystyle \mathrm {H} _{n}} should be maximal if all the outcomes are equally likely i.e.Hn(p1,…,pn)≤Hn(1n,…,1n){\displaystyle \mathrm {H} _{n}(p_{1},\ldots ,p_{n})\leq \mathrm {H} _{n}\left({\frac {1}{n}},\ldots ,{\frac {1}{n}}\right)}.
  4. Increasing number of outcomes: for equiprobable events, the entropy should increase with the number of outcomes i.e.Hn(1n,…,1n⏟n)<Hn+1(1n+1,…,1n+1⏟n+1).{\displaystyle \mathrm {H} _{n}{\bigg (}\underbrace {{\frac {1}{n}},\ldots ,{\frac {1}{n}}} _{n}{\bigg )}<\mathrm {H} _{n+1}{\bigg (}\underbrace {{\frac {1}{n+1}},\ldots ,{\frac {1}{n+1}}} _{n+1}{\bigg )}.}
  5. Additivity: given an ensemble ofn uniformly distributed elements that are partitioned intok boxes (sub-systems) withb1, ...,bk elements each, the entropy of the whole ensemble should be equal to the sum of the entropy of the system of boxes and the individual entropies of the boxes, each weighted with the probability of being in that particular box.

Discussion

[edit]

The rule of additivity has the following consequences: forpositive integersbi whereb1 + ... +bk =n,Hn(1n,…,1n)=Hk(b1n,…,bkn)+∑i=1kbinHbi(1bi,…,1bi).{\displaystyle \mathrm {H} _{n}\left({\frac {1}{n}},\ldots ,{\frac {1}{n}}\right)=\mathrm {H} _{k}\left({\frac {b_{1}}{n}},\ldots ,{\frac {b_{k}}{n}}\right)+\sum _{i=1}^{k}{\frac {b_{i}}{n}}\,\mathrm {H} _{b_{i}}\left({\frac {1}{b_{i}}},\ldots ,{\frac {1}{b_{i}}}\right).}

Choosingk =n,b1 = ... =bn = 1 this implies that the entropy of a certain outcome is zero:Η1(1) = 0. This implies that the efficiency of a source set withn symbols can be defined simply as being equal to itsn-ary entropy. See alsoRedundancy (information theory).

The characterization here imposes an additive property with respect to apartition of a set. Meanwhile, theconditional probability is defined in terms of a multiplicative property,P(A∣B)⋅P(B)=P(A∩B){\displaystyle P(A\mid B)\cdot P(B)=P(A\cap B)}. Observe that a logarithm mediates between these two operations. Theconditional entropy and related quantities inherit simple relation, in turn. The measure theoretic definition in the previous section defined the entropy as a sum over expected surprisalsμ(A)⋅ln⁡μ(A){\displaystyle \mu (A)\cdot \ln \mu (A)} for an extremal partition. Here the logarithm is ad hoc and the entropy is not a measure in itself. At least in the information theory of a binary string,log2{\displaystyle \log _{2}} lends itself to practical interpretations.

Motivated by such relations, a plethora of related and competing quantities have been defined. For example,David Ellerman's analysis of a "logic of partitions" defines a competing measure in structuresdual to that of subsets of a universal set.[14] Information is quantified as "dits" (distinctions), a measure on partitions. "Dits" can be converted intoShannon's bits, to get the formulas for conditional entropy, and so on.

Alternative characterization via additivity and subadditivity

[edit]

Another succinct axiomatic characterization of Shannon entropy was given byAczél, Forte and Ng,[15] via the following properties:

  1. Subadditivity:H(X,Y)≤H(X)+H(Y){\displaystyle \mathrm {H} (X,Y)\leq \mathrm {H} (X)+\mathrm {H} (Y)} for jointly distributed random variablesX,Y{\displaystyle X,Y}.
  2. Additivity:H(X,Y)=H(X)+H(Y){\displaystyle \mathrm {H} (X,Y)=\mathrm {H} (X)+\mathrm {H} (Y)} when the random variablesX,Y{\displaystyle X,Y} are independent.
  3. Expansibility:Hn+1(p1,…,pn,0)=Hn(p1,…,pn){\displaystyle \mathrm {H} _{n+1}(p_{1},\ldots ,p_{n},0)=\mathrm {H} _{n}(p_{1},\ldots ,p_{n})}, i.e., adding an outcome with probability zero does not change the entropy.
  4. Symmetry:Hn(p1,…,pn){\displaystyle \mathrm {H} _{n}(p_{1},\ldots ,p_{n})} is invariant under permutation ofp1,…,pn{\displaystyle p_{1},\ldots ,p_{n}}.
  5. Small for small probabilities:limq→0+H2(1−q,q)=0{\displaystyle \lim _{q\to 0^{+}}\mathrm {H} _{2}(1-q,q)=0}.

Discussion

[edit]

It was shown that any functionH{\displaystyle \mathrm {H} } satisfying the above properties must be a constant multiple of Shannon entropy, with a non-negative constant.[15] Compared to the previously mentioned characterizations of entropy, this characterization focuses on the properties of entropy as a function of random variables (subadditivity and additivity), rather than the properties of entropy as a function of the probability vectorp1,…,pn{\displaystyle p_{1},\ldots ,p_{n}}.

It is worth noting that if we drop the "small for small probabilities" property, thenH{\displaystyle \mathrm {H} } must be a non-negative linear combination of the Shannon entropy and theHartley entropy.[15]

Further properties

[edit]

The Shannon entropy satisfies the following properties, for some of which it is useful to interpret entropy as the expected amount of information learned (or uncertainty eliminated) by revealing the value of a random variableX:

Aspects

[edit]

Relationship to thermodynamic entropy

[edit]
Main article:Entropy in thermodynamics and information theory

The inspiration for adopting the wordentropy in information theory came from the close resemblance between Shannon's formula and very similar known formulae fromstatistical mechanics.

Instatistical thermodynamics the most general formula for the thermodynamicentropyS of athermodynamic system is theGibbs entropyS=−kB∑ipiln⁡pi,{\displaystyle S=-k_{\text{B}}\sum _{i}p_{i}\ln p_{i}\,,}wherekB is theBoltzmann constant, andpi is the probability of amicrostate. TheGibbs entropy was defined byJ. Willard Gibbs in 1878 after earlier work byLudwig Boltzmann (1872).[16]

The Gibbs entropy translates over almost unchanged into the world ofquantum physics to give thevon Neumann entropy introduced byJohn von Neumann in 1927:S=−kBTr(ρln⁡ρ),{\displaystyle S=-k_{\text{B}}\,{\rm {Tr}}(\rho \ln \rho )\,,}where ρ is thedensity matrix of the quantum mechanical system and Tr is thetrace.[17]

At an everyday practical level, the links between information entropy and thermodynamic entropy are not evident. Physicists and chemists are apt to be more interested inchanges in entropy as a system spontaneously evolves away from its initial conditions, in accordance with thesecond law of thermodynamics, rather than an unchanging probability distribution. As the minuteness of the Boltzmann constantkB indicates, the changes inS /kB for even tiny amounts of substances in chemical and physical processes represent amounts of entropy that are extremely large compared to anything indata compression orsignal processing. In classical thermodynamics, entropy is defined in terms of macroscopic measurements and makes no reference to any probability distribution, which is central to the definition of information entropy.

The connection between thermodynamics and what is now known as information theory was first made by Boltzmann and expressed by hisequation:

S=kBln⁡W,{\displaystyle S=k_{\text{B}}\ln W,}

whereS{\displaystyle S} is the thermodynamic entropy of a particular macrostate (defined by thermodynamic parameters such as temperature, volume, energy, etc.),W is the number of microstates (various combinations of particles in various energy states) that can yield the given macrostate, andkB is the Boltzmann constant.[18] It is assumed that each microstate is equally likely, so that the probability of a given microstate ispi = 1/W. When these probabilities are substituted into the above expression for the Gibbs entropy (or equivalentlykB times the Shannon entropy), Boltzmann's equation results. In information theoretic terms, the information entropy of a system is the amount of "missing" information needed to determine a microstate, given the macrostate.

In the view ofJaynes (1957),[19] thermodynamic entropy, as explained bystatistical mechanics, should be seen as anapplication of Shannon's information theory: the thermodynamic entropy is interpreted as being proportional to the amount of further Shannon information needed to define the detailed microscopic state of the system, that remains uncommunicated by a description solely in terms of the macroscopic variables of classical thermodynamics, with the constant of proportionality being just the Boltzmann constant. Adding heat to a system increases its thermodynamic entropy because it increases the number of possible microscopic states of the system that are consistent with the measurable values of its macroscopic variables, making any complete state description longer. (See article:maximum entropy thermodynamics).Maxwell's demon can (hypothetically) reduce the thermodynamic entropy of a system by using information about the states of individual molecules; but, asLandauer (from 1961) and co-workers[20] have shown, to function the demon himself must increase thermodynamic entropy in the process, by at least the amount of Shannon information he proposes to first acquire and store; and so the total thermodynamic entropy does not decrease (which resolves the paradox).Landauer's principle imposes a lower bound on the amount of heat a computer must generate to process a given amount of information, though modern computers are far less efficient.

Data compression

[edit]
Main articles:Shannon's source coding theorem andData compression

Shannon's definition of entropy, when applied to an information source, can determine the minimum channel capacity required to reliably transmit the source as encoded binary digits. Shannon's entropy measures the information contained in a message as opposed to the portion of the message that is determined (or predictable). Examples of the latter include redundancy in language structure or statistical properties relating to the occurrence frequencies of letter or word pairs, triplets etc. The minimum channel capacity can be realized in theory by using thetypical set or in practice usingHuffman,Lempel–Ziv orarithmetic coding. (See alsoKolmogorov complexity.) In practice, compression algorithms deliberately include some judicious redundancy in the form ofchecksums to protect against errors. Theentropy rate of a data source is the average number of bits per symbol needed to encode it. Shannon's experiments with human predictors show an information rate between 0.6 and 1.3 bits per character in English;[21] thePPM compression algorithm can achieve a compression ratio of 1.5 bits per character in English text.

If acompression scheme is lossless – one in which you can always recover the entire original message by decompression – then a compressed message has the same quantity of information as the original but is communicated in fewer characters. It has more information (higher entropy) per character. A compressed message has lessredundancy.Shannon's source coding theorem states a lossless compression scheme cannot compress messages, on average, to havemore than one bit of information per bit of message, but that any valueless than one bit of information per bit of message can be attained by employing a suitable coding scheme. The entropy of a message per bit multiplied by the length of that message is a measure of how much total information the message contains. Shannon's theorem also implies that no lossless compression scheme can shortenall messages. If some messages come out shorter, at least one must come out longer due to thepigeonhole principle. In practical use, this is generally not a problem, because one is usually only interested in compressing certain types of messages, such as a document in English, as opposed to gibberish text, or digital photographs rather than noise, and it is unimportant if a compression algorithm makes some unlikely or uninteresting sequences larger.

A 2011 study inScience estimates the world's technological capacity to store and communicate optimally compressed information normalized on the most effective compression algorithms available in the year 2007, therefore estimating the entropy of the technologically available sources.[22]: 60–65 

All figures in entropically compressedexabytes
Type of Information19862007
Storage2.6295
Broadcast4321900
Telecommunications0.28165

The authors estimate humankind technological capacity to store information (fully entropically compressed) in 1986 and again in 2007. They break the information into three categories—to store information on a medium, to receive information through one-waybroadcast networks, or to exchange information through two-waytelecommunications networks.[22]

Entropy as a measure of diversity

[edit]
Main article:Diversity index

Entropy is one of several ways to measure biodiversity and is applied in the form of theShannon index.[23] A diversity index is a quantitative statistical measure of how many different types exist in a dataset, such as species in a community, accounting for ecologicalrichness,evenness, anddominance. Specifically, Shannon entropy is the logarithm of1D, thetrue diversity index with parameter equal to 1. The Shannon index is related to the proportional abundances of types.

Entropy of a sequence

[edit]

There are a number of entropy-related concepts that mathematically quantify information content of a sequence or message:

  • theself-information of an individual message or symbol taken from a given probability distribution (message or sequence seen as an individual event),
  • thejoint entropy of the symbols forming the message or sequence (seen as a set of events),
  • theentropy rate of astochastic process (message or sequence is seen as a succession of events).

(The "rate of self-information" can also be defined for a particular sequence of messages or symbols generated by a given stochastic process: this will always be equal to the entropy rate in the case of astationary process.) Otherquantities of information are also used to compare or relate different sources of information.

It is important not to confuse the above concepts. Often it is only clear from context which one is meant. For example, when someone says that the "entropy" of the English language is about 1 bit per character, they are actually modeling the English language as a stochastic process and talking about its entropyrate. Shannon himself used the term in this way.

If very large blocks are used, the estimate of per-character entropy rate may become artificially low because the probability distribution of the sequence is not known exactly; it is only an estimate. If one considers the text of every book ever published as a sequence, with each symbol being the text of a complete book, and if there areN published books, and each book is only published once, the estimate of the probability of each book is1/N, and the entropy (in bits) is−log2(1/N) = log2(N). As a practical code, this corresponds to assigning each book aunique identifier and using it in place of the text of the book whenever one wants to refer to the book. This is enormously useful for talking about books, but it is not so useful for characterizing the information content of an individual book, or of language in general: it is not possible to reconstruct the book from its identifier without knowing the probability distribution, that is, the complete text of all the books. The key idea is that the complexity of the probabilistic model must be considered.Kolmogorov complexity is a theoretical generalization of this idea that allows the consideration of the information content of a sequence independent of any particular probability model; it considers the shortestprogram for auniversal computer that outputs the sequence. A code that achieves the entropy rate of a sequence for a given model, plus the codebook (i.e. the probabilistic model), is one such program, but it may not be the shortest.

The Fibonacci sequence is 1, 1, 2, 3, 5, 8, 13, .... treating the sequence as a message and each number as a symbol, there are almost as many symbols as there are characters in the message, giving an entropy of approximatelylog2(n). The first 128 symbols of the Fibonacci sequence has an entropy of approximately 7 bits/symbol, but the sequence can be expressed using a formula [F(n) = F(n−1) + F(n−2) forn = 3, 4, 5, ...,F(1) =1,F(2) = 1] and this formula has a much lower entropy and applies to any length of the Fibonacci sequence.

Limitations of entropy in cryptography

[edit]

Incryptanalysis, entropy is often roughly used as a measure of the unpredictability of a cryptographic key, though its realuncertainty is unmeasurable. For example, a 128-bit key that is uniformly and randomly generated has 128 bits of entropy. It also takes (on average)2127{\displaystyle 2^{127}} guesses to break by brute force. Entropy fails to capture the number of guesses required if the possible keys are not chosen uniformly.[24][25] Instead, a measure calledguesswork can be used to measure the effort required for a brute force attack.[26]

Other problems may arise from non-uniform distributions used in cryptography. For example, a 1,000,000-digit binaryone-time pad using exclusive or. If the pad has 1,000,000 bits of entropy, it is perfect. If the pad has 999,999 bits of entropy, evenly distributed (each individual bit of the pad having 0.999999 bits of entropy) it may provide good security. But if the pad has 999,999 bits of entropy, where the first bit is fixed and the remaining 999,999 bits are perfectly random, the first bit of the ciphertext will not be encrypted at all.

Data as a Markov process

[edit]

A common way to define entropy for text is based on theMarkov model of text. For an order-0 source (each character is selected independent of the last characters), the binary entropy is:

H(S)=−∑ipilog⁡pi,{\displaystyle \mathrm {H} ({\mathcal {S}})=-\sum _{i}p_{i}\log p_{i},}

wherepi is the probability ofi. For a first-orderMarkov source (one in which the probability of selecting a character is dependent only on the immediately preceding character), theentropy rate is:[citation needed]

H(S)=−∑ipi∑j pi(j)log⁡pi(j),{\displaystyle \mathrm {H} ({\mathcal {S}})=-\sum _{i}p_{i}\sum _{j}\ p_{i}(j)\log p_{i}(j),}

wherei is astate (certain preceding characters) andpi(j){\displaystyle p_{i}(j)} is the probability ofj giveni as the previous character.

For a second order Markov source, the entropy rate is

H(S)=−∑ipi∑jpi(j)∑kpi,j(k) log⁡pi,j(k).{\displaystyle \mathrm {H} ({\mathcal {S}})=-\sum _{i}p_{i}\sum _{j}p_{i}(j)\sum _{k}p_{i,j}(k)\ \log p_{i,j}(k).}

Efficiency (normalized entropy)

[edit]

A source setX{\displaystyle {\mathcal {X}}} with a non-uniform distribution will have less entropy than the same set with a uniform distribution (i.e. the "optimized alphabet"). This deficiency in entropy can be expressed as a ratio called efficiency:[27]

η(X)=HHmax=−∑i=1np(xi)logb⁡(p(xi))logb⁡(n).{\displaystyle \eta (X)={\frac {H}{H_{\text{max}}}}=-\sum _{i=1}^{n}{\frac {p(x_{i})\log _{b}(p(x_{i}))}{\log _{b}(n)}}.}Applying the basic properties of the logarithm, this quantity can also be expressed as:η(X)=−∑i=1np(xi)logb⁡(p(xi))logb⁡(n)=∑i=1nlogb⁡(p(xi)−p(xi))logb⁡(n)=∑i=1nlogn⁡(p(xi)−p(xi))=logn⁡(∏i=1np(xi)−p(xi)).{\displaystyle {\begin{aligned}\eta (X)&=-\sum _{i=1}^{n}{\frac {p(x_{i})\log _{b}(p(x_{i}))}{\log _{b}(n)}}=\sum _{i=1}^{n}{\frac {\log _{b}\left(p(x_{i})^{-p(x_{i})}\right)}{\log _{b}(n)}}\\[1ex]&=\sum _{i=1}^{n}\log _{n}\left(p(x_{i})^{-p(x_{i})}\right)=\log _{n}\left(\prod _{i=1}^{n}p(x_{i})^{-p(x_{i})}\right).\end{aligned}}}

Efficiency has utility in quantifying the effective use of acommunication channel. This formulation is also referred to as the normalized entropy, as the entropy is divided by the maximum entropylogb⁡(n){\displaystyle {\log _{b}(n)}}. Furthermore, the efficiency is indifferent to the choice of (positive) baseb, as indicated by the insensitivity within the final logarithm above thereto.

Entropy for continuous random variables

[edit]

Differential entropy

[edit]
Main article:Differential entropy

The Shannon entropy is restricted to random variables taking discrete values. The corresponding formula for a continuous random variable withprobability density functionf(x) with finite or infinite supportX{\displaystyle \mathbb {X} } on the real line is defined by analogy, using the above form of the entropy as an expectation:[10]: 224 

H(X)=E[−log⁡f(X)]=−∫Xf(x)log⁡f(x)dx.{\displaystyle \mathrm {H} (X)=\mathbb {E} [-\log f(X)]=-\int _{\mathbb {X} }f(x)\log f(x)\,\mathrm {d} x.}

This is the differential entropy (or continuous entropy). A precursor of the continuous entropyh[f] is the expression for the functionalΗ in theH-theorem of Boltzmann.

Although the analogy between both functions is suggestive, the following question must be set: is the differential entropy a valid extension of the Shannon discrete entropy? Differential entropy lacks a number of properties that the Shannon discrete entropy has – it can even be negative – and corrections have been suggested, notablylimiting density of discrete points.

To answer this question, a connection must be established between the two functions:

In order to obtain a generally finite measure as thebin size goes to zero. In the discrete case, the bin size is the (implicit) width of each of then (finite or infinite) bins whose probabilities are denoted bypn. As the continuous domain is generalized, the width must be made explicit.

To do this, start with a continuous functionf discretized into bins of sizeΔ{\displaystyle \Delta }.By the mean-value theorem there exists a valuexi in each bin such thatf(xi)Δ=∫iΔ(i+1)Δf(x)dx{\displaystyle f(x_{i})\Delta =\int _{i\Delta }^{(i+1)\Delta }f(x)\,dx}the integral of the functionf can be approximated (in the Riemannian sense) by∫−∞∞f(x)dx=limΔ→0∑i=−∞∞f(xi)Δ,{\displaystyle \int _{-\infty }^{\infty }f(x)\,dx=\lim _{\Delta \to 0}\sum _{i=-\infty }^{\infty }f(x_{i})\Delta ,}where this limit and "bin size goes to zero" are equivalent.

We will denoteHΔ:=−∑i=−∞∞f(xi)Δlog⁡(f(xi)Δ){\displaystyle \mathrm {H} ^{\Delta }:=-\sum _{i=-\infty }^{\infty }f(x_{i})\Delta \log \left(f(x_{i})\Delta \right)}and expanding the logarithm, we haveHΔ=−∑i=−∞∞f(xi)Δlog⁡(f(xi))−∑i=−∞∞f(xi)Δlog⁡(Δ).{\displaystyle \mathrm {H} ^{\Delta }=-\sum _{i=-\infty }^{\infty }f(x_{i})\Delta \log(f(x_{i}))-\sum _{i=-\infty }^{\infty }f(x_{i})\Delta \log(\Delta ).}

AsΔ → 0, we have

∑i=−∞∞f(xi)Δ→∫−∞∞f(x)dx=1∑i=−∞∞f(xi)Δlog⁡(f(xi))→∫−∞∞f(x)log⁡f(x)dx.{\displaystyle {\begin{aligned}\sum _{i=-\infty }^{\infty }f(x_{i})\Delta &\to \int _{-\infty }^{\infty }f(x)\,dx=1\\\sum _{i=-\infty }^{\infty }f(x_{i})\Delta \log(f(x_{i}))&\to \int _{-\infty }^{\infty }f(x)\log f(x)\,dx.\end{aligned}}}

Note;log(Δ) → −∞ asΔ → 0, requires a special definition of the differential or continuous entropy:

h[f]=limΔ→0(HΔ+log⁡Δ)=−∫−∞∞f(x)log⁡f(x)dx,{\displaystyle h[f]=\lim _{\Delta \to 0}\left(\mathrm {H} ^{\Delta }+\log \Delta \right)=-\int _{-\infty }^{\infty }f(x)\log f(x)\,dx,}

which is, as said before, referred to as the differential entropy. This means that the differential entropyis not a limit of the Shannon entropy forn → ∞. Rather, it differs from the limit of the Shannon entropy by an infinite offset (see also the article oninformation dimension).

Limiting density of discrete points

[edit]
Main article:Limiting density of discrete points

It turns out as a result that, unlike the Shannon entropy, the differential entropy isnot in general a good measure of uncertainty or information. For example, the differential entropy can be negative; also it is not invariant under continuous co-ordinate transformations. This problem may be illustrated by a change of units whenx is a dimensioned variable.f(x) will then have the units of1/x. The argument of the logarithm must be dimensionless, otherwise it is improper, so that the differential entropy as given above will be improper. IfΔ is some "standard" value ofx (i.e. "bin size") and therefore has the same units, then a modified differential entropy may be written in proper form as:H=∫−∞∞f(x)log⁡(f(x)Δ)dx,{\displaystyle \mathrm {H} =\int _{-\infty }^{\infty }f(x)\log(f(x)\,\Delta )\,dx,}and the result will be the same for any choice of units forx. In fact, the limit of discrete entropy asN→∞{\displaystyle N\rightarrow \infty } would also include a term oflog⁡(N){\displaystyle \log(N)}, which would in general be infinite. This is expected: continuous variables would typically have infinite entropy when discretized. Thelimiting density of discrete points is really a measure of how much easier a distribution is to describe than a distribution that is uniform over its quantization scheme.

Relative entropy

[edit]
Main article:Generalized relative entropy

Another useful measure of entropy that works equally well in the discrete and the continuous case is therelative entropy of a distribution. It is defined as theKullback–Leibler divergence from the distribution to a reference measurem as follows. Assume that a probability distributionp isabsolutely continuous with respect to a measurem, i.e. is of the formp(dx) =f(x)m(dx) for some non-negativem-integrable functionf withm-integral 1, then the relative entropy can be defined asDKL(p‖m)=∫log⁡(f(x))p(dx)=∫f(x)log⁡(f(x))m(dx).{\displaystyle D_{\mathrm {KL} }(p\|m)=\int \log(f(x))p(dx)=\int f(x)\log(f(x))m(dx).}

In this form the relative entropy generalizes (up to change in sign) both the discrete entropy, where the measurem is thecounting measure, and the differential entropy, where the measurem is theLebesgue measure. If the measurem is itself a probability distribution, the relative entropy is non-negative, and zero ifp =m as measures. It is defined for any measure space, hence coordinate independent and invariant under co-ordinate reparameterizations if one properly takes into account the transformation of the measurem. The relative entropy, and (implicitly) entropy and differential entropy, do depend on the "reference" measurem.

Use in number theory

[edit]

Terence Tao used entropy to make a useful connection trying to solve theErdős discrepancy problem.[28][29]

Intuitively the idea behind the proof was if there is low information in terms of the Shannon entropy between consecutive random variables (here the random variable is defined using theLiouville function (which is a useful mathematical function for studying distribution of primes)XH =λ(n+H){\displaystyle \lambda (n+H)}. And in an interval [n, n+H] the sum over that interval could become arbitrary large. For example, a sequence of +1's (which are values ofXH could take) have trivially low entropy and their sum would become big. But the key insight was showing a reduction in entropy by non negligible amounts as one expands H leading inturn to unbounded growth of a mathematical object over this random variable is equivalent to showing the unbounded growth per theErdős discrepancy problem.

The proof is quite involved and it brought together breakthroughs not just in novel use of Shannon entropy, but also it used theLiouville function along with averages of modulated multiplicative functions[30] in short intervals. Proving it also broke the "parity barrier"[31] for this specific problem.

While the use of Shannon entropy in the proof is novel it is likely to open new research in this direction.

Use in combinatorics

[edit]

Entropy has become a useful quantity incombinatorics.

Loomis–Whitney inequality

[edit]

A simple example of this is an alternative proof of theLoomis–Whitney inequality: for every subsetA ⊆Zd, we have|A|d−1≤∏i=1d|Pi(A)|{\displaystyle |A|^{d-1}\leq \prod _{i=1}^{d}|P_{i}(A)|}wherePi is theorthogonal projection in theith coordinate:Pi(A)={(x1,…,xi−1,xi+1,…,xd):(x1,…,xd)∈A}.{\displaystyle P_{i}(A)=\{(x_{1},\ldots ,x_{i-1},x_{i+1},\ldots ,x_{d}):(x_{1},\ldots ,x_{d})\in A\}.}

The proof follows as a simple corollary ofShearer's inequality: ifX1, ...,Xd are random variables andS1, ...,Sn are subsets of{1, ...,d} such that every integer between 1 andd lies in exactlyr of these subsets, thenH[(X1,…,Xd)]≤1r∑i=1nH[(Xj)j∈Si]{\displaystyle \mathrm {H} [(X_{1},\ldots ,X_{d})]\leq {\frac {1}{r}}\sum _{i=1}^{n}\mathrm {H} [(X_{j})_{j\in S_{i}}]}where(Xj)j∈Si{\displaystyle (X_{j})_{j\in S_{i}}} is the Cartesian product of random variablesXj with indexesj inSi (so the dimension of this vector is equal to the size ofSi).

We sketch how Loomis–Whitney follows from this: Indeed, letX be a uniformly distributed random variable with values inA and so that each point inA occurs with equal probability. Then (by the further properties of entropy mentioned above)Η(X) = log|A|, where|A| denotes the cardinality ofA. LetSi = {1, 2, ...,i−1,i+1, ...,d}. The range of(Xj)j∈Si{\displaystyle (X_{j})_{j\in S_{i}}} is contained inPi(A) and henceH[(Xj)j∈Si]≤log⁡|Pi(A)|{\displaystyle \mathrm {H} [(X_{j})_{j\in S_{i}}]\leq \log |P_{i}(A)|}. Now use this to bound the right side of Shearer's inequality and exponentiate the opposite sides of the resulting inequality you obtain.

Approximation to binomial coefficient

[edit]

For integers0 <k <n letq =k/n. Then2nH(q)n+1≤(nk)≤2nH(q),{\displaystyle {\frac {2^{n\mathrm {H} (q)}}{n+1}}\leq {\tbinom {n}{k}}\leq 2^{n\mathrm {H} (q)},}where[32]: 43 H(q)=−qlog2⁡(q)−(1−q)log2⁡(1−q).{\displaystyle \mathrm {H} (q)=-q\log _{2}(q)-(1-q)\log _{2}(1-q).}

Proof (sketch)
Note that(nk)qqn(1−q)n−nq{\displaystyle {\tbinom {n}{k}}q^{qn}(1-q)^{n-nq}} is one term of the expression

∑i=0n(ni)qi(1−q)n−i=(q+(1−q))n=1.{\displaystyle \sum _{i=0}^{n}{\tbinom {n}{i}}q^{i}(1-q)^{n-i}=(q+(1-q))^{n}=1.} Rearranging gives the upper bound. For the lower bound one first shows, using some algebra, that it is the largest term in the summation. But then,(nk)qqn(1−q)n−nq≥1n+1{\displaystyle {\binom {n}{k}}q^{qn}(1-q)^{n-nq}\geq {\frac {1}{n+1}}}since there aren + 1 terms in the summation. Rearranging gives the lower bound.

A nice interpretation of this is that the number of binary strings of lengthn with exactlyk many 1's is approximately2nH(k/n){\displaystyle 2^{n\mathrm {H} (k/n)}}.[33]

Use in machine learning

[edit]

Machine learning techniques arise largely from statistics and also information theory. In general, entropy is a measure of uncertainty and the objective of machine learning is to minimize uncertainty.

Decision tree learning algorithms use relative entropy to determine the decision rules that govern the data at each node.[34] Theinformation gain in decision treesIG(Y,X){\displaystyle IG(Y,X)}, which is equal to the difference between the entropy ofY{\displaystyle Y} and the conditional entropy ofY{\displaystyle Y} givenX{\displaystyle X}, quantifies the expected information, or the reduction in entropy, from additionally knowing the value of an attributeX{\displaystyle X}. The information gain is used to identify which attributes of the dataset provide the most information and should be used to split the nodes of the tree optimally.

Bayesian inference models often apply theprinciple of maximum entropy to obtainprior probability distributions.[35] The idea is that the distribution that best represents the current state of knowledge of a system is the one with the largest entropy, and is therefore suitable to be the prior.

Classification in machine learning performed bylogistic regression orartificial neural networks often employs a standard loss function, calledcross-entropy loss, that minimizes the average cross entropy between ground truth and predicted distributions.[36] In general, cross entropy is a measure of the differences between two datasets similar to the KL divergence (also known as relative entropy).

See also

[edit]

Notes

[edit]
  1. ^This definition allows events with probability 0, resulting in the undefinedlog⁡(0){\displaystyle \log(0)}. We do seelimx→0xlog⁡(x)=0{\displaystyle \lim \limits _{x\rightarrow 0}x\log(x)=0} and it can be assumed that0log⁡(0){\displaystyle 0\log(0)} equals 0 in this context. Alternatively one can definep:X→(0,1]{\displaystyle p\colon {\mathcal {X}}\to (0,1]}, not allowing events with probability equal to exactly 0.

References

[edit]
  1. ^Pathria, R. K.; Beale, Paul (2011).Statistical Mechanics (Third ed.). Academic Press. p. 51.ISBN 978-0123821881.
  2. ^abShannon, Claude E. (July 1948)."A Mathematical Theory of Communication".Bell System Technical Journal.27 (3):379–423.Bibcode:1948BSTJ...27..379S.doi:10.1002/j.1538-7305.1948.tb01338.x.hdl:10338.dmlcz/101429. (PDF, archived fromhereArchived 20 June 2014 at theWayback Machine)
  3. ^abShannon, Claude E. (October 1948)."A Mathematical Theory of Communication".Bell System Technical Journal.27 (4):623–656.Bibcode:1948BSTJ...27..623S.doi:10.1002/j.1538-7305.1948.tb00917.x.hdl:11858/00-001M-0000-002C-4317-B. (PDF, archived fromhereArchived 10 May 2013 at theWayback Machine)
  4. ^"Entropy (for data science) Clearly Explained!!!". 24 August 2021.Archived from the original on 5 October 2021. Retrieved5 October 2021 – viaYouTube.
  5. ^MacKay, David J.C. (2003).Information Theory, Inference, and Learning Algorithms. Cambridge University Press.ISBN 0-521-64298-1.Archived from the original on 17 February 2016. Retrieved9 June 2014.
  6. ^Shannon, Claude Elwood; Weaver, Warren (1998).The mathematical theory of communication. Urbana: Univ. of Illinois Press. p. 15.ISBN 978-0-252-72548-7.
  7. ^Schneier, B:Applied Cryptography, Second edition, John Wiley and Sons.
  8. ^Borda, Monica (2011).Fundamentals in Information Theory and Coding. Springer.ISBN 978-3-642-20346-6.
  9. ^Han, Te Sun; Kobayashi, Kingo (2002).Mathematics of Information and Coding. American Mathematical Society.ISBN 978-0-8218-4256-0.
  10. ^abcdefghijkThomas M. Cover; Joy A. Thomas (1991).Elements of Information Theory. Hoboken, New Jersey: Wiley.ISBN 978-0-471-24195-9.
  11. ^Entropy at thenLab
  12. ^Carter, Tom (March 2014).An introduction to information theory and entropy(PDF). Santa Fe.Archived(PDF) from the original on 4 June 2016. Retrieved4 August 2017.{{cite book}}: CS1 maint: location missing publisher (link)
  13. ^Chakrabarti, C. G., and Indranil Chakrabarty. "Shannon entropy: axiomatic characterization and application."International Journal of Mathematics and Mathematical Sciences 2005. 17 (2005): 2847-2854urlArchived 5 October 2021 at theWayback Machine
  14. ^Ellerman, David (October 2017)."Logical Information Theory: New Logical Foundations for Information Theory"(PDF).Logic Journal of the IGPL.25 (5):806–835.doi:10.1093/jigpal/jzx022.Archived(PDF) from the original on 25 December 2022. Retrieved2 November 2022.
  15. ^abcAczél, J.; Forte, B.; Ng, C. T. (1974). "Why the Shannon and Hartley entropies are 'natural'".Advances in Applied Probability.6 (1):131–146.doi:10.2307/1426210.JSTOR 1426210.S2CID 204177762.
  16. ^Compare: Boltzmann, Ludwig (1896, 1898). Vorlesungen über Gastheorie : 2 Volumes – Leipzig 1895/98 UB: O 5262-6. English version: Lectures on gas theory. Translated by Stephen G. Brush (1964) Berkeley: University of California Press; (1995) New York: DoverISBN 0-486-68455-5
  17. ^Życzkowski, Karol (2006).Geometry of Quantum States: An Introduction to Quantum Entanglement. Cambridge University Press. p. 301.
  18. ^Sharp, Kim; Matschinsky, Franz (2015)."Translation of Ludwig Boltzmann's Paper "On the Relationship between the Second Fundamental Theorem of the Mechanical Theory of Heat and Probability Calculations Regarding the Conditions for Thermal Equilibrium"".Entropy.17:1971–2009.doi:10.3390/e17041971.
  19. ^Jaynes, E. T. (15 May 1957)."Information Theory and Statistical Mechanics".Physical Review.106 (4):620–630.Bibcode:1957PhRv..106..620J.doi:10.1103/PhysRev.106.620.S2CID 17870175.
  20. ^Landauer, R. (July 1961)."Irreversibility and Heat Generation in the Computing Process".IBM Journal of Research and Development.5 (3):183–191.doi:10.1147/rd.53.0183.ISSN 0018-8646.Archived from the original on 15 December 2021. Retrieved15 December 2021.
  21. ^Mark Nelson (24 August 2006)."The Hutter Prize". Archived fromthe original on 1 March 2018. Retrieved27 November 2008.
  22. ^ab"The World's Technological Capacity to Store, Communicate, and Compute Information"Archived 27 July 2013 at theWayback Machine, Martin Hilbert and Priscila López (2011),Science, 332(6025); free access to the article through here: martinhilbert.net/WorldInfoCapacity.html
  23. ^Spellerberg, Ian F.; Fedor, Peter J. (2003)."A tribute to Claude Shannon (1916–2001) and a plea for more rigorous use of species richness, species diversity and the 'Shannon–Wiener' Index".Global Ecology and Biogeography.12 (3):177–179.Bibcode:2003GloEB..12..177S.doi:10.1046/j.1466-822X.2003.00015.x.ISSN 1466-8238.S2CID 85935463.
  24. ^Massey, James (1994)."Guessing and Entropy"(PDF).Proc. IEEE International Symposium on Information Theory.Archived(PDF) from the original on 1 January 2014. Retrieved31 December 2013.
  25. ^Malone, David; Sullivan, Wayne (2005)."Guesswork is not a Substitute for Entropy"(PDF).Proceedings of the Information Technology & Telecommunications Conference.Archived(PDF) from the original on 15 April 2016. Retrieved31 December 2013.
  26. ^Pliam, John (1999). "Selected Areas in Cryptography".International Workshop on Selected Areas in Cryptography. Lecture Notes in Computer Science. Vol. 1758. pp. 62–77.doi:10.1007/3-540-46513-8_5.ISBN 978-3-540-67185-5.
  27. ^Indices of Qualitative Variation.AR Wilcox - 1967https://www.osti.gov/servlets/purl/4167340
  28. ^Klarreich, Erica (1 October 2015)."A Magical Answer to an 80-Year-Old Puzzle".Quanta Magazine. Retrieved18 August 2014.
  29. ^Tao, Terence (28 February 2016)."The Erdős discrepancy problem".Discrete Analysis.arXiv:1509.05363v6.doi:10.19086/da.609.S2CID 59361755.Archived from the original on 25 September 2023. Retrieved20 September 2023.
  30. ^https://arxiv.org/pdf/1502.02374.pdfArchived 28 October 2023 at theWayback Machine
  31. ^https://terrytao.wordpress.com/2007/06/05/open-question-the-parity-problem-in-sieve-theory/Archived 7 August 2023 at theWayback Machine
  32. ^Aoki, New Approaches to Macroeconomic Modeling.
  33. ^Probability and Computing, M. Mitzenmacher and E. Upfal, Cambridge University Press
  34. ^Batra, Mridula; Agrawal, Rashmi (2018)."Comparative Analysis of Decision Tree Algorithms". In Panigrahi, Bijaya Ketan; Hoda, M. N.; Sharma, Vinod; Goel, Shivendra (eds.).Nature Inspired Computing. Advances in Intelligent Systems and Computing. Vol. 652. Singapore: Springer. pp. 31–36.doi:10.1007/978-981-10-6747-1_4.ISBN 978-981-10-6747-1.Archived from the original on 19 December 2022. Retrieved16 December 2021.
  35. ^Jaynes, Edwin T. (September 1968). "Prior Probabilities".IEEE Transactions on Systems Science and Cybernetics.4 (3):227–241.Bibcode:1968IJSSC...4..227J.doi:10.1109/TSSC.1968.300117.ISSN 2168-2887.
  36. ^Rubinstein, Reuven Y.; Kroese, Dirk P. (9 March 2013).The Cross-Entropy Method: A Unified Approach to Combinatorial Optimization, Monte-Carlo Simulation and Machine Learning. Springer Science & Business Media.ISBN 978-1-4757-4321-0.

This article incorporates material from Shannon's entropy onPlanetMath, which is licensed under theCreative Commons Attribution/Share-Alike License.

Further reading

[edit]

Textbooks on information theory

[edit]

External links

[edit]
Library resources about
Entropy (information theory)
Lossless
type
Entropy
Dictionary
Other
Hybrid
Lossy
type
Transform
Predictive
Audio
Concepts
Codec
parts
Image
Concepts
Methods
Video
Concepts
Codec
parts
Theory
Community
People
International
National
Other
Retrieved from "https://en.wikipedia.org/w/index.php?title=Entropy_(information_theory)&oldid=1322644264"
Categories:
Hidden categories:
[8]ページ先頭
©2009-2025 Movatter.jp