CIA-RDP96-00787R000500070003-9

Approved For Release 2000/08/07 : CIA-RDP96-00787R000500070003-9
SG11
Using table 4 and de 80 probability
being anest that they write on page Ily thay
following can be stated
Thereare 32 possible 5 fit permentions
for generated outpute. all 32 can be grouped
with into a 0, 1 a 2 enn manage
box with a .94208 reliability (Some of the
peramestimations fit into foot 1 and 2 enon bake
and all of t 2 enn entries an
watch to book he (00000 and the 11011) on
A (01110 and 10101) correct message.
That is
2 eur.
wages de not (in PAT auger)
uniquely identify a connect message, but only
narrow de libaly choices x 2,
Because of the 4 correct message oder den
Love, sevent thing Regen. If a crest massage
i received, (eq. 00000) Then it must be never be. ..
Approved For Release 2000/08/07 : CIA-RDP96-00787R0005008/0003-9
3
med, 01/10 0 10101

Approved For Release 2000/08/07 : CIA-RDP96-00787R000500070003-9
2
4 anos message / 1011
Simel
. O ena message (p == 32768) is roughly
6 times in likely an a 3 error message (P=0.05/2)
amal 50 times an litalien de a 4 euros message (.0064)
den ch 00000 knie in et leur 6 times
(p =. 82) more likely han each of it. the choices
but in only about 3 times ( p. = . 6) |.0512+0512
+, 0064 = 1088| more likely them all
ter messages. Most of de a 1 eva messages
have even smaller added unrobot & dan
it uniquely identifying a specific correct
message, For instance, the receipt of 00100 i
Can shown in Table 4) a one error message / 00000
However ( mot chon in Table 4) i also a 2 evan
managed fox 01110 and 10101. Since
at olle of a two over message and p: 2048 and
to probilities, the message 00100 the
hoan ~ 15 change of the message 00000
Chiny either 011on 10101)
0- 4096 relation & p= 4096 fer (, 2048+ 2048
Approved"i or Helease 2000/0810/ : CIA-RDP96-00787R000500070003-9

Approved For Release 2000/08/07 : CIA-RDP96-00787R000500070003-9
The statistics get hairy here and A
Loven + tried to do them fully but the
curve for "no error correcting" is very wrong
( see minted green curve to a Table 4) and
I would bet that the "over correcting" cove is
also enmeously high, House, & can inspiron
of the accuracy of the " majority vote" curve.
SG11
-
Approved For Release 2000/08/07 : CIA-RDP96-00787R000500070003-9

P,(00000) : 100128
Pc = 175
Approved For Release 2000/08/07 : CIA-RDP96-00787R009500070003-9
01/10
10/01
E= . 02688
11011
P3 (10101) = . 00512
P2 (11011) : . 02048
~ 3:9
p = (8)5
32768
(.B) (2)
E.08192
PA ( 8) ( 2)
= 02048
P = (a) (2)
.00212
P = (8) (2) = 100 $28
P5 (2)5
2 .00036
1
P (00000) =
. 00512
(13.5%)
-
(01110) =,02048
(37,5 %)
12
(37.5%)
P2 (roii) =
P3 (11011) =
(12,5%)
Approved For Release 2000/08/07 : CIA-RDP96-00787R000500070003-9