Difference between revisions of "Aufgaben:Exercise 1.5Z: Symmetrical Markov Source"

From LNTwww
 
(One intermediate revision by the same user not shown)
Line 1: Line 1:
  
{{quiz-Header|Buchseite=Informationstheorie/Nachrichtenquellen mit Gedächtnis
+
{{quiz-Header|Buchseite=Information_Theory/Discrete_Sources_with_Memory
 
}}
 
}}
  
Line 130: Line 130:
 
*Now neither  $\rm A$  can directly follow  $\rm A$  nor  $\rm B$  can directly follow  $\rm B$ .  
 
*Now neither  $\rm A$  can directly follow  $\rm A$  nor  $\rm B$  can directly follow  $\rm B$ .  
 
*The result is always an alternating sequence, depending on the starting value the sequence  $\rm ABABAB$ ...  or  $\rm BABABA$... .  
 
*The result is always an alternating sequence, depending on the starting value the sequence  $\rm ABABAB$ ...  or  $\rm BABABA$... .  
*This source also has the entropy  $H = H_{\rm bin}(1) = 0$ in both cases.
+
*This source also has the entropy  $H = H_{\rm bin}(1) = 0$  in both cases.
 
{{ML-Fuß}}
 
{{ML-Fuß}}
  

Latest revision as of 13:04, 10 August 2021

Binary symmetrical Markov diagram

Exercise 1.5  dealt with a binary Markov source in which the transition probabilities from  $\rm A$  to  $\rm B$  and from  $\rm B$  to  $\rm A$  were different.

In this task, the following shall now apply:

$$p_{\rm A\hspace{0.01cm}|\hspace{0.01cm}B} = p_{\rm B\hspace{0.01cm}|\hspace{0.01cm}A} = q \hspace{0.8cm} ( 0 \le q \le 1) \hspace{0.05cm}.$$

All equations given in Exercise 1.5 also apply here:

  • Entropy:
$$H = p_{\rm AA} \cdot {\rm log_2}\hspace{0.1cm}\frac {1}{ p_{\rm A\hspace{0.01cm}|\hspace{0.01cm}A}} + p_{\rm AB} \cdot {\rm log_2}\hspace{0.1cm}\frac {1}{ p_{\rm B\hspace{0.01cm}|\hspace{0.01cm}A}} + p_{\rm BA} \cdot {\rm log_2}\hspace{0.1cm}\frac {1}{ p_{\rm A\hspace{0.01cm}|\hspace{0.01cm}B}} + p_{\rm BB} \cdot {\rm log_2}\hspace{0.1cm}\frac {1}{ p_{\rm B\hspace{0.01cm}|\hspace{0.01cm}B}} \hspace{0.05cm}.$$
  • First entropy approximation::
$$H_{\rm 1} = p_{\rm A} \cdot {\rm log_2}\hspace{0.1cm} \frac{1}{p_{\rm A}} + p_{\rm B} \cdot {\rm log_2}\hspace{0.1cm} \frac{1}{p_{\rm B}} \hspace{0.05cm}.$$
  • k–th entropy approximation $(k = 2, 3, \ \text{...})$:
$$H_k = {1}/{k} \cdot \big [ H_{\rm 1} + (k-1) \cdot H \big] \hspace{0.05cm},\hspace{0.5cm}H = \lim_{k \rightarrow \infty } H_k \hspace{0.05cm}.$$




Hints:



Questions

1

Calculate the symbol probabilities for the transition probability  $q = 1/4$.

$p_{\rm A} \ = \ $

$p_{\rm B} \ = \ $

2

Calculate the source entropy  $H$  for  $q = 1/4$.

$H \ =$

$\ \rm bit/symbol$

3

What entropy approximations are obtained for  $q = 1/4$?

$H_1 \ = \ $

$\ \rm bit/symbol$
$H_2 \ = \ $

$\ \rm bit/symbol$
$H_3 \ = \ $

$\ \rm bit/symbol$

4

Determine  $q$  such that  $H$  becomes maximum. Interpretation.

$q \ = \ $

5

Which symbol sequences are possible with  $q = 0$ ?

$\rm AAAAAA$ ...
$\rm BBBBBB$ ...
$\rm ABABAB$ ...

6

Which symbol sequences are possible with  $q = 1$ ?

$\rm AAAAAA$ ...
$\rm BBBBBB$ ...
$\rm ABABAB$ ...


Solution

(1)  For a stationary first order binary Markov source holds:

$$p_{\rm A} = p_{\rm A\hspace{0.01cm}|\hspace{0.01cm}A} \cdot p_{\rm A} + p_{\rm A\hspace{0.01cm}|\hspace{0.01cm}B} \cdot p_{\rm B} = (1-q) \cdot p_{\rm A} + q \cdot p_{\rm B}$$
$$\Rightarrow \hspace{0.3cm}q \cdot p_{\rm A} = q \cdot p_{\rm B} \hspace{0.3cm} \Rightarrow \hspace{0.3cm}p_{\rm A} = p_{\rm B}\hspace{0.15cm} \underline {= 0.5} \hspace{0.05cm}.$$


(2)  To calculate the entropy  $H$  one needs all four composite probabilities:

$$p_{\rm AA} \hspace{0.1cm} = \hspace{0.1cm} p_{\rm A} \cdot p_{\rm A\hspace{0.01cm}|\hspace{0.01cm}A} = 1/2 \cdot(1-q) = p_{\rm BB}\hspace{0.05cm},\hspace{1cm} p_{\rm AB} \hspace{0.1cm} = \hspace{0.1cm} p_{\rm A} \cdot p_{\rm B\hspace{0.01cm}|\hspace{0.01cm}A} = 1/2 \cdot q = p_{\rm BA}\hspace{0.05cm}.$$
  • Substituting these values into the given entropy equation, we get
$$H = 2 \cdot \frac{1}{2} \cdot(1-q) \cdot {\rm log}_2\hspace{0.1cm} \frac{1}{1-q} + 2 \cdot \frac{1}{2} \cdot q \cdot {\rm log}_2\hspace{0.1cm} \frac{1}{q} = q \cdot {\rm log}_2\hspace{0.1cm} \frac{1}{q} + (1-q) \cdot {\rm log}_2\hspace{0.1cm} \frac{1}{1-q} = H_{\rm bin}(q) \hspace{0.05cm}.$$
  • The numerical value sought is  $H = H_{\rm bin} (0.25) \hspace{0.15cm}\underline{= 0.811 \, \rm bit/symbol}$.


(3)  For equally probable binary symbols,  $H_1 \hspace{0.15cm}\underline{= 1 \, \rm bit/Symbol}$. 

  • Using the equation valid for Markov sources, it further holds:
$$H_2 \hspace{0.1cm} = \hspace{0.1cm} {1}/{2} \cdot \big[ H_1 + H \big] \hspace{0.15cm} \underline {= 0.906 \,{\rm bit/Symbol}} \hspace{0.05cm},$$
$$ H_3 \hspace{0.1cm} = \hspace{0.1cm} {1}/{3} \cdot \big[ H_1 + 2 H \big] \hspace{0.15cm} \underline {= 0.874 \,{\rm bit/Symbol}} \hspace{0.05cm}.$$


(4)  The maximum of the binary entropy function is obtained for  $q\hspace{0.15cm}\underline{= 0.5}$.

  • Thus the maximum entropy is  $H = 1 \, \rm bit/symbol$.
  • It can be seen from the relationship  $H = H_1$  and from the transition diagram shown at the front that  $q = 0.5$  results in statistically independent symbols:
$$p_{\rm A} = p_{\rm A\hspace{0.01cm}|\hspace{0.01cm}A} = p_{\rm A\hspace{0.01cm}|\hspace{0.01cm}B} = 0.5 \hspace{0.05cm}, \hspace{0.2cm} p_{\rm B} = p_{\rm B\hspace{0.01cm}|\hspace{0.01cm}A} = p_{\rm B\hspace{0.01cm}|\hspace{0.01cm}B}= 0.5 \hspace{0.05cm}.$$


(5)  Proposed solutions 1 and 2 are correct:

  • The symbol sequence results either in  $\rm AAAAAA$ ...  or in  $\rm BBBBBB$ ... , depending on which symbol was given as the starting value.
  • The entropy of such a source is always  $H = H_{\rm bin}(0) = 0$.



(6)  Only proposed solution 3 is correct:

  • Now neither  $\rm A$  can directly follow  $\rm A$  nor  $\rm B$  can directly follow  $\rm B$ .
  • The result is always an alternating sequence, depending on the starting value the sequence  $\rm ABABAB$ ...  or  $\rm BABABA$... .
  • This source also has the entropy  $H = H_{\rm bin}(1) = 0$  in both cases.