Information Technology Reference
In-Depth Information
FIGURE 5. Probability for drawing a white ball at the
n
th trial from an urn having
initially four balls of which 1, 2, or 3 are white, the others black. White balls are
replaced, black are not (a). Entropy at the
n
th trial (b).
which is the solution of
dp
dn
(
)
=
q 1
pp
-
(43)
or, recursively expressed, of
()
=-
(
)
+
(
)
◊
(
)
pn
pn
1
q
pn
-
1
qn
-
1
(44)
Figure 5a compares the probabilities
p
(
n
) for drawing a white ball at
the
n
th
trial, as calculated through approximation [Eq. (42)] (solid curves),
with the exact values computed by an IBM 360/50 system with a pro-
gram kindly supplied by Mr. Atwood for an urn with initially four balls
(
N
= 4) and for the three cases in which one, two, or three of these are
white (
N
w
= 1,
N
w
= 2,
N
w
= 3). The entropy*
H
(
n
) in bits per trial corre-
sponding to these cases is shown in Fig. 5b, and one may note that while
for some cases [
p
(1) £ 0.5] it reaches a maximum in the course of this game,
* Or the “amount of uncertainty”, or the “amount of information” received by the
outcome of each trial, defined by -H(
n
) =
p
(
n
)log
2
p
(
n
) +
q
(
n
)log
2
q
(
n
).
Search WWH ::
Custom Search