25.08.2015 Views

In the Beginning was Information

6KezkB

6KezkB

SHOW MORE
SHOW LESS
  • No tags were found...

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

To compute <strong>the</strong> average information content per symbol I ave , we have todivide <strong>the</strong> number given by equation (6) by <strong>the</strong> number of symbols concerned:NI ave = I tot /n = ∑ p(x i ) x lb(1/p(x i )). (9)i=1When equation (9) is evaluated for <strong>the</strong> frequencies of <strong>the</strong> letters occurringin English, <strong>the</strong> values shown in Table 1, are obtained. The average informationcontent of one letter is I ave = 4.045 77. The corresponding value forGerman is I ave = 4.112 95.The average I ave (x) which can be computed from equation (9) thus is <strong>the</strong>arithmetic mean of <strong>the</strong> all <strong>the</strong> single values I(x). The average informationcontent of every symbol is given in Table 1 for two different symbol systems(<strong>the</strong> English and German alphabets); for <strong>the</strong> sake of simplicity i is usedinstead of I ave . The average information content for each symbol I ave (x) ≡ i is<strong>the</strong> same as <strong>the</strong> expectation value 22 of <strong>the</strong> information content of one symbolin a long sequence. This quantity is also known as <strong>the</strong> entropy 23 H of <strong>the</strong>source of <strong>the</strong> message or of <strong>the</strong> employed language (I ave ≡ i ≡ H). Equation(9) is a fundamental expression in Shannon’s <strong>the</strong>ory. It can be interpretedin various ways:a) <strong>In</strong>formation content of each symbol: H is <strong>the</strong> average information contentI ave (x) of a symbol x i in a long sequence of n symbols. H thus is acharacteristic of a language when n is large enough. Because of <strong>the</strong> differentletter frequencies in various languages, H has a specific value for everylanguage (e. g. H 1 = 4.045 77 for English and for German it is 4.112 95).22 Expectation value: The expectation value E is a concept which is defined for randomquantities in probability calculus. The sum ∑ p k x g(x k ) taken over all k singlevalues, is called <strong>the</strong> expectation value E of <strong>the</strong> probability distribution, where g(x) isa given discrete distribution with x k as abscissae and p k as ordinates (= <strong>the</strong> probabilityof appearance of <strong>the</strong> values x k ). This value is also known as <strong>the</strong> mean value or <strong>the</strong>ma<strong>the</strong>matical hope.23 Entropy: This concept <strong>was</strong> first introduced in <strong>the</strong>rmodynamics by Rudolf Clausiusabout 1850. Later, in 1877, Ludwig Boltzmann (1844-1906) showed that entropy isproportional to <strong>the</strong> logarithm of <strong>the</strong> probability of a system being in a certain state.Because <strong>the</strong> formal derivation of <strong>the</strong> ma<strong>the</strong>matical formulas for physical entropy issimilar to equation (9), Shannon (1948) also called this quantity entropy. Unfortunately<strong>the</strong> use of <strong>the</strong> same term for such fundamentally different phenomena hasresulted in many erroneous conclusions. When <strong>the</strong> second law of <strong>the</strong>rmodynamics,which is also known as <strong>the</strong> entropy <strong>the</strong>orem, is flippantly applied to Shannon’s informationconcept, it only causes confusion. <strong>In</strong> <strong>the</strong>rmodynamics entropy depends on<strong>the</strong> temperature, which can not at all be said of informational entropy.175

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!