Information Technology Reference
In-Depth Information
iteration, its AL should be decreased. Module i is able to influence the environment only if
its AL exceeds a threshold (p5) at iteration m, then
k
k
(
IV
CV
)
,
im
,
im
,
im
im
OV
otherwise
OV
0
(7)
,
,
z
The environment is also a n-dimension vector (E) and its k
th
component at iteration m is
changed by the following formula:
iz
,
im
OV
1
k
k
i
1
Ec s
tan
t
ise
(8)
m
m
z
It is not difficult to see that, unlike the analysis of neural mechanisms above, this model
does not include special layers to form output. The objective of such a design is to avoid
unnecessary difficulties conditioned by complex relations between such layers. These
difficulties are able to complicate the understanding of the model's functioning without
clearing its main ideas. However, because each module has a complex structure with
internal vectors such as CV and OV the model can be useful to understand the functioning
of various goal-directed systems.
In all simulations, the number of modules in the system (z) was 300 and the vectors in each
module were three-dimensional. Real numbers were used as the stuff of all vectors in the
system. A goal-directed activity was constructed as follows. First, 40 modules were
considered as stable processors. Its coefficient vectors were filled by a constant plus small
noise and its active levels was more than p5 (0.3). The other modules of the system were
flexible elements. Its coefficient vectors were randomly filled by numbers from 0 to 100 and
its ALs were randomly established at 0,06 plus small noise. Following this initialization,
interactions between stable processors and flexible elements started. Five iterations of this
process took place and p1 was 0,3. At this stage (stage 1), no outputs from the system
influenced the environment. This corresponded to the construction of a goal-directed
activity. After this, novel constants were established and the interaction between the system
and the environment became possible. This stage (stage 2) meant the functioning of a goal-
directed activity.
It is necessary to emphasize that the architecture of the model means stable processors are
not a necessary condition for the interaction between the system and the environment. In
principle, flexible elements are sufficient to provide the functioning of the system but in this
case, the activity of the system must be less stable and persistent. To test this suggestion a
special simulation was carried out. In this simulation no stable processors were formed but
five iterations similar to those in simulation 1 were performed (stage 1, accordingly). After
this, certain constants were selected and the interaction between the system and the
environment became possible (stage 2).
In both simulations all constants were 80, in other words, the goal of the activity in
simulation 1 entirely met the initial state of its stable processors. After three iterations with
such constants at stage 2, in both simulations the constants were forcefully established at 50
during one iteration to estimate the stability of the models to random fluctuations.
