Professional Documents
Culture Documents
Hans Marko The Bidirectional Communication Theroy - A Generalization of Information Theory 1973
Hans Marko The Bidirectional Communication Theroy - A Generalization of Information Theory 1973
Hans Marko The Bidirectional Communication Theroy - A Generalization of Information Theory 1973
COMMUNICATIONS,
ON
and Europzan areas where he designed and installed LF and HF audio teletype systems. He
engaged in signal andcomponent design for
UHF and microwave scatter and satellite communications systems. From1967to1969he
was a manager of .the ALTAIR radar site at
Kwajalein. Since 1970, as the Assistant Leader
of the Educational TechnologyGroup atLincoln
Laboratory, Massachusetts Institute of Technology, Lexington,he has workedon the Lincoln
Training System.
I. INTRODUCTION
1346
Forthispurpose
the following essential characteristics are
considered.
1) Information transmitter and receiver of a human being
are identical; a person has a characteristic information source
which is differentfromtheinformation
source ofanother
person. Therefore a mansupplies a stochastic processwhich
is typical for him with an entropy which characterizes him.
2) Information transmission betweentwo persons is considered as a !stochastic synchronization of thisstochastic
process. The entropy of each process has therefore a free
and a dependent part.
The latter represents the received
transinformation.
To describe this idea mathematically, the theory of Markov
processes is used;and stationarity has tobe assumed.
Even the philosophers of antiquity tried to describe the
process of thinking bymeans of association andmemory.
Aristotle described with a remarkable clearness the ,sequence
of imagination as a statistical process with inner linkage, as we
would call it liowadays, in his short essay Memory and
Recollection. According to him, recollection is characterized
by the fact that movements (i.e., imagination) follow each
otherhabitually
(i.e., mostly).The
presentsequence
is
essentially determined by the sequence of the earlier process.
Caused by the mechanical systems of Galilei andNewton, the
psychology of association came to a concept of a mechanisticdeterministic behavior in its search fora physics of soul.
Locke
and
Hartley
in
England and Herbart in Germany
attributedthe process of thinkingtoa causally determined
mechanism of association.This view has been corrected by
the modern psychology of thinking which recognized the importanceofintuition.Naturally,the
value of all of these
theories is limited because the statements cannot be quantized.
At theendofthe19thcentury,Galton,
Ebbinghaus,and
Wundt began with the experimentalinvestigation of association
processes, including association sequences. Finally, Shannons
fundamental work [ 11 rendered a quantitative description and
therefore made ameasuring of information possible.
The idea ofthe bidirectional communicationtheory was
first presentedbytheauthoratthecybernetics
congress at
Kiel, Germany, in September 1965 [2] and in May 1966 with
a lecture at the congress of the Popov Society in Moscow. An
explicit representation of this theory isgiven in thejournal
Kybernetik [ 3 ] , and short representations are given in [4] and
[ 5 ] . Related mathematicalproofs are presented in [6] and
[7]. An extension to the communication of a group has been
given byNeuburger [8], [9]. Thebidirectionalcommunication theory has beenapplied so far with behavioral sciences
[IO], [ 111 . Other applications for statistically coupled
systems, i.e., economical systems,are possible.
Two-way communication channels have beeninvestigated
earlier by Shannon and others, especially the feedback channel
[ 12]., [ 131 and the crosstalk channel [ 141 . In these investigations, however, the conventional definitionof transinformation
is used, and the information source is considered independent
and not statistically dependent, as in the present work.
The same applies for previous work with the aim to investigatemultivariate
correlation using informationalquantities
Fig. 1 shows the block diagram of unidirectionaltransmission according to Shannon. It consists of a message source
a coding device (C) which codes the message in an appro:
priate way for the transmission, the transmission channel,a
and the receiver (R). The channel (Ch)
decoding device (D),
contains a noise source ( N ) which represents the disturbance.
It is essential foraquantitativedescription
of information
transmission. Thus the block diagram contains two statistical
generators: the message source Q andthe noise source N .
According to the usual symbolism, the sources are represented
by circles; all the other parts are passive and drawn as boxes.
The receiver is passive in contrast to the bidirectional communication model. The receiver usually is considered as ideal,
supposing thatit canevaluate the received message in its
statistical propertiesoptimally (i.e., it is supposed to have
storage ofinfinite size). Because of these assumptions,the
transmissionbecomes independent of the receiver and is determined solely by the properties of the message sourceand
thechannel.Toapply
Shannons formulasforthe
general
case of a channel with memory considered here, stationarity
must be assumed. A sequence of symbols x, = x l x 2 . . . at the
receiver is observed. Both transmitted and received sequences
are supposed to have n symbols. The followingprobabilities
are defined.
(e),
p(x,)
Probability
for
the
occurrence
of
the
sequence
x, at the transmitter.
pb,)
Probability
for
the
occurrence
of
the
sequence
receiver.
p ( x , y,) Joint probability for the occurrence of x , and y,.
p(x,ly,) Conditionalprobability.for
t h e occurrenceof
x, when y , is known.
p b , [x,) Conditionalprobabilityfortheoccurrenceof
y , when x
, is known.
.
yn the
at
{ } designates theexpectation value (mean value). Theentropies related to onesymbol can be calculated fromthe
probabilities as follows.
Entropy at the transmitter:
1
H ( x ) = lim - { - log p ( x , ) } .
n-+m
1347
source
coder
decoder
rqwrer
channel
Ch
! y
1 noire
channel.
-+
1348
Directed transinformation M1
--f
M2 :
Coincidence M1 +-+ M 2 :
The last case (dialogue) is the most general: the first two
cases are contained in it as special cases. Thedefinitions of
the following part therefore refer to this general case.
K =-TI2-I-T21.
P(YIYnXn) = ~ ( Y l ~ m x m )
are valid for Markov processes of the order m when n 2 m .
The transition probabilities p(xl xmym)and p(ylymxm)
determine the two stochastic processes completely; therefore
they can be designated as generator probabilities.
The symbol { } represents expectation values. The following equations represent mean information values per symbol
(entropy), according to the following definitions.
Total information ofMI :
(7)
Free information o f M l :
F 1 = lim { - logp(xlx,y,)}.
n+
(8)
Hi
= Ti2
H2
+ Fl
T21 -t F2.
R 1 = H 1 - K =F1
R2
(14)
=Hz
T21
- K = F 2 - T12.
-
(15)
(1 6 )
called residual
(17)
(18)
Then the above relations can be represented clearly by the information flow diagram of Fig. 4. The two stochastical
generators of MI and M 2 generate thefreeinformation. At
the nodes (15)-( 18) are valid as Kirchhoff laws.
1349
T12
F1 + F z = R I + R , + K = H 1+ H z K
- .
(19)
From this we get an important statement. The larger the coincidence, that is, thetotaltransmittedtransinformation
in
both directions, the smaller is (for given entropies H 1 and H z )
the sum of the free informations. An effective communication
limits the sum of the free generated information, which seems
to be logical because of thestrong coupling of the .two
processes in this case. There is another limitation for the magnitude of the transinformation flows whichresults from the
identity of thetotalinformation defined herewith Shannons
entropy. The following inequalities are valid, since Shannons
transinformation, here the coincidence, is always less than
H(x) as well as H ( y ) .
H1 & K
Hz
2K.
(20)
R1
t
Rl=Hl
Tl2 = F2
(26)
T21 =F1
(27)
(21)
1350
M1
'G1
,suggestion M2 +Ml
171
and 172.
Subdominod
onimd
"dictator"-
<
Tl2
T2 1
= 1.
T12 + T21 T21 + T12
tehaviour
'
ACKNOWLEDGMENT
The author wishes to thank Dr. Neuburger, who helped with
many fruitful discussions and mathematical proofs, and who
extended the theory to the multidirectionalcase.
REFERENCES
[ 11 C. E. Shannon, A mathematicaltheory
of communication,
Bell Syst. Tech. J., vol. 21, pp. 379-423, 623-652, 1948.
[2] H. Marko, Informationstheorie und Kybernetik.Fortschr.d.Kybernetik, in Bericht uber die Tagung Kiel 1965 der Deutschen
Arbeitsgemeinschaft
Kybernetik.
Munich, Germany: Oldenbourg-Verlag, 1967, pp. 9-28.
[ 31 -,
Die Theorie der bidirektionalen Kommunikation und ihre
Anwendung auf die Informationsiibermittlung zwischen Menschen
(subjective information), Kybernetik, vol. 3, pp. 128-136, 1966.
Information theory and cybernetics, IEEESpectrum,
[4] -,
pp. 75-83,1967.
[5] H. Marko and E. Neuburger, A short reviewof thetheory of
bidirectionalcommunication, Nachrichtentech. Z., vol. 6,pp.
320-323, 1970.
[6] -, Uber gerichtete Groi3en in der Informationstheorie. Untersuchungen zur Theoriederbidirektionalen
Kommunikation,
Arch. Elek. Ubertragung, vol. 21, no. 2, pp. 61-69, 1967.
[ 71 E. Neuburger, Zwei Fundamentalgesetzederbidirektionalen
Kommunikation, Arch. Elek. Ubertragung, no. 5, pp. 208-214,
1970.
Beschreibung der gegenseitigen Abhangigkeit von Signal[8] -,
folgen durchgerichteteInformationsgroDen,
Nachrichtentech.
Fachberichte, vol. 33, pp. 49-59, 1967.
[9] -, Kommunikation der Gruppe (Ein Beitrag zur Informationstheorie). Munich, Germany: Oldenbourg-Verlag, 1970.
1351