Information Technology Reference
In-Depth Information
exterior reward. At the balance condition, all rules in the chain have the same
intensity. At the unbalance condition, all rules in the chain are described by
(13.31), not by (13.32). In any case, if expected reward received by rule chain is
well consistent, then the intensity of the rules in the chain will converge to
general level.
Table 13.2 Different methods to modify intensity
Ruel
PSP's intensity
BBA's intensity
R
1
R
2
R
3
R
4
R
5
R
6
R
7
1000
299
1000
4
300
999
300
648
567
645
644
300
531
300
To compare PSP with BBA, let's consider Figure 13.11. This figure presents
10 states, including initial states
A
and
B
, terminal states
H
,
I
and
J
. The exterior
rewards generated in states
are 1000, 0 and 300, respectively.
In this example, classifier system uses PSP or BBA to modify the intensity of
rule. All initial intensities are set to be 100, and bidding rate b is 0.1. In addition,
interior rewards are distributed to 1000 plots. This means that each classifier
system consists of 3000 steps. The obtained intensities of rules are presented in
Table 13.2.
H
,
I
and
J
,QLWLDOVWDWH
%
$
5
5
'
&
5
5
5
(
)
*
5
5
5
+
,
-
5HZDUG
7HUPLQDOVWDWH
Fig. 13.11. Comparison of PSP and BBA
