Você está na página 1de 6

A SHORT DISCUSSION ON THE PARADOX.

COUNTERPOINT BETWEEN
CLASSICAL LOGIC AND COMPENSATORY FUZZY LOGIC

Solving a true paradox concludes with no actual conclusion


Fernando Otero

Department of Mathematics, Faculty of Engineering, National University of Mar del Plata, JB Justo 4302,
Mar del Plata, Buenos Aires, Argentina


ABSTRACT

This is a brief discussion on the notion of paradox comparing some results from different perspectives of
logic, i.e. classical Boolean logic and compensatory fuzzy logic. We synthetize our analysis by examining
a practical example.

INTRODUCTION

Paradox is extended to different contexts and applies to different kind of problems. However we want to
focus on a brief analysis of the logical interpretation. In terms of classical Boolean propositional logic, a
paradox appears when a contradiction on a logical value of a Boolean variable. For instance, you assume
some proposition to be false and that assumption leads you to its truth and viceversa, you assume it to be
true and you conclude with the falsity of the proposition. A paradox can be represented in many different
equivalent logical terms. The most simple manner to express a paradox is denoted in eq.(1) using
implication operator =>, conjunction operator .and negation operator ~ in the traditional propositional
logic:

( ) ( ) p p p p . ~ ~ (1)

or in a resumed form such as eq.(2)

p p ~
(2)

It is important to notice that both implications or both ways of implication are needed in order to get a
paradox. One way of implication is correct, that is in Boolean logic from a false premise you can reach a
true conclusion following a correct implication. The other direction is actually forbidden. You can never
conclude a false statement since a true one. We can give a subtle example of this in classical words of Lao
Tse:

To realize that our knowledge is ignorance is a noble insight. To regard our
ignorance as knowledge is mental sickness.


DISCUSSION

In this summarized paper we intend to examine the concept of paradox in a dynamic manner, that means
analyzing evolution of a paradox in time such as a simple form of eq.(3)

) ( ~ ) 1 ( t p t p = +
(3)



In digital logic circuitry, we have several types of dynamic paradoxes of the form of eq.(3). We have
Ring Oscillators, composed of an odd number of NOT gates whose output oscillates between two voltage
levels. This final output is asserted a finite amount of time after the first input is asserted; the feedback of
this last output to the input causes oscillation. The oscillator period is in all cases equal to twice the sum
of the individual delays of all stages. A typical example of this type of oscillator is in figure 1.


Figure 1. A schematic of a simple 3-inverter ring oscillator whose frequency is 1/(6*inverter delay)

It is important to notice that this configuration of inverter feedback can be used as a storage element; it is
the basic building block of static random access memory, or SRAM. Ring oscillators are so basic and
important in digital logic hardware that they are often used to demonstrate a new hardware technology,
analogous to the way a hello world program is often used to demonstrate a new software technology

Another typical example of digital logic circuitry using the concept of dynamic paradox is using an
universal logical gate as the NOR gate as seen in fig. 2.



Figure 2. Schematic Circuit for an oscillator built from a NOR gate


These two simple examples as we have roughly seen, show how paradoxical digital logic allows creation
of an oscillating logic (we excluded other typical effects observed in digital oscillators such as jitter for
reasons of simplicity). However we will see in the Compensatory Fuzzy Logic (CFL), how this
oscillating logic, obtained as a result of classical Boolean logic can be turned into a damped response.

First, lets define in a simple manner what we mean by the CFL: in basic terms, we will refer CFL to
some fuzzy logic model that allows the simultaneous modeling of the deductive and decision-making
processes (Cejas and Montero, 2011) based on the removal of some axioms in order to achieve a sensitive
and idempotent multi-valued system (Bouchet et al., 2011)










Now lets re-express the dynamic form of the paradox of eq. (3) in the form of an iterative process which
can be modeled as

.... )) ( ( ) (
0 0 0
x f f x f x
(4)

where
) (
0
x f
is the fuzzy operation corresponding to the implication
p p ~
starting from the
inicial value
0
x .

At this point, the CFL may bring us several definitions fur the fuzzy operation of eq.(4), and even the
implication
p p ~
has two possible definitions (Bouchet et al, 2010), remembering that
y x
can be expressed as:

y x y x v ~
(5)
) ( ~ y x x y x . v
(6)

Even when both definitions have the same Boolean logic table, we will see that according to the CFL they
will behave very differently.

Well focus on the CFL based on the geometrical mean and on the arithmetical mean with the
corresponding generalized definitions of conjunction, disjunction and negation (Bouchet et al., 2010):

CFL based on the geometrical mean

n
n n
x x x x x x
/ 1
2 1 2 1
) .... . ( ... . . . (7.1)
| |
n
n n
x x x x x x
/ 1
2 1 2 1
) 1 )....( 1 ).( 1 ( 1 ... v v v (7.2)
1 1
1 ~ x x (7.3)

CFL based on the arithmetical mean

2 / 1
1
2 1 2 1
) ,..., , min( ...
(
(
(
(

. . .

=
n
x
x x x x x x
n
i
i
n n
(8.1)

2 / 1
1
2 1 2 1
1
) 1 ,..., 1 , 1 min( 1 ...
(
(
(
(

v v v

=
n
x
x x x x x x
n
i
i
n n
(8.2)


1 1
1 ~ x x
(8.3)





For both cases of the CFL described before, according to definition 1), we have obtained an oscillating
function:

0 0
1 ) ( x x f =
(9)

in consistency with the oscillating behavior showed in the Boolean logic.

However , using the definition 2) gives some different results:

After some straightforward algebra we arrive at eqns. (10.1) and (10.2) for ) (
0
x f for the CFL based on
the geometrical mean and based on the arithmetical mean respectively:

| | ) 1 ( 1 1 ) (
0 0 0 0
x x x x f =
(10.1)


| | | |
|
|
.
|

\
|

|
|
.
|

\
|
+ = ) 1 ( , min
2
1
1 , min ) 1 ( , min
2
1
1
2
1
1 ) (
0 0 0 0 0 0 0
x x x x x x x f

(10.2)

So by analyzing the iterative form of eq.(4) for CFL using definition 2) starting from different initial
values, we conclude that only for extreme truth values 0 and 1, the behaviour is a continuous oscillation
exactly the same as in the classical Boolean case, but for any other intermediate values, iterations
converge to 0.5. Lets compare the iterative response for eqns. (9), (10.1) and (10.2) for several initial
values.

0 5 10 15 20 25 30
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1

Figure 3. Responses for CFL for def. 1) (blue full line), for def. 2) based on geometrical mean (dotted red line)
and based on arithmetical mean (segmented green line), all of them starting from a logical value of 0.99


0 5 10 15 20 25 30
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7

Figure 4. Responses for CFL for def. 1) (blue full line), for def. 2) based on geometrical mean (dotted red line)
and based on arithmetical mean (segmented green line), all of them starting from a logical value of 0.6754


In fig. 3, we started from a near to 1 logic value ( 99 . 0
0
= x ) and at iteration 10, CFL based on
geometrical mean arrives at 0.5005, while CFL based on arithmetic mean arrives at 0.4825, showing a
relative faster performance on convergence in the first case. In fig. 4, we started from a value
6754 . 0
0
= x and at iteration 10, CFL based on geometrical mean arrives at 0.4997, while CFL based on
arithmetic mean arrives at 0.4971, in this case with similar performance of both types of CFL, but still
faster convergence for the CFL based on geometrical mean. In these two examples convergence achieves
a similar order of precision with a similar number of iterations. Necessary iterations to achieve certain
order of precision depend on how far from the 0.5 are initial values and how precise are these initial
values.

The overdamped response of eqns. (10.1) and (10.2) which converges to 0.5 may be thought as an
iterative process of running along a Moebius strip as in fig. 5 and each iteration corresponds to a
displacement. If you take as logical states, being in the inner and the outer side, as 0 and 1 respectively, it
is reasonable to conclude that it's neither true nor false to be in any side. This is particularly true for the
example of the Moebius strip taking account that it is a non-orientable surface.


Figure 5. Sculptural representation of a Moebius strip.





CONCLUSIONS

Paradoxes defined in a dynamic form are traditionally related to an oscillating response, however as we
generalize Boolean logic to Compensatory Fuzzy Logic (CFL) and apply a second definition for the
logical operation of implication, results change. A convergence to a state of truth neither true nor false is
reached in some iterations when we start from any other value than 0 and 1. We compare results applying
CFL both based on geometrical mean and based on arithmetical mean, where the first one seems to have a
faster convergence. CFL seems appropriate to work with paradoxes, deleting (in some manner) classical
difficulties that appear in classical Boolean logic.


ACKNOWLEDGEMENTS

The author wants to thank Dr. Peter Breuer at the University of Birmingham, for a interesting discussion
about the topic.

REFERENCES

1. Cejas-Montero, J., La lgica difusa compensatoria / The compensatory fuzzy logic, Ingeniera
Industrial, Vol. 32 (2), pp. 157, 2011
2. Bouchet, A., Pastore, J. I., Espn Andrade R., Brun, M. and Ballarn, V., Arithmetic mean based
compensatory fuzzy logic, Journal of Computational Intelligence and Applications, Vol. 10 (2), pp. 231,
2011
3. Bouchet, A., Pastore, J., Brun, M. and Ballarn, V., Lgica Difusa Compensatoria basada en la media
aritmtica y su aplicacin en la Morfologa Matemtica Difusa, IV Torneo Regional de Inteligencia
Computacional (TRIC 2010), Buenos Aires, 2010

Você também pode gostar