Information Technology Reference
In-Depth Information
These include: the method of agent's learning from its own mistakes and the
method of agent's learning through imitation. Obviously, these ways of learning do
not use all possibilities of the agent's learning and other methods of learning, which
are popular and developed in the field of artificial intelligence, may also be used
successfully.
The agent's process of learning based on its mistakes may be described with the
use of a scenario realized in the following stages:
•
The agent chooses the optimal strategy
s
based on the comparison between the
models
m
and
m
and realizes this strategy in the surrounding environment with
the use of the operation
X
(Sect.
3.4
).
After realizing this strategy the agent builds the new model
m
based on the
observation of the surrounding environment.
•
•
The agent compares the models of the environment, the one it wanted to receive
(
m
) and the one which was created in reality (
m
).
If the difference between the models
m
and
m
is too great (above a certain
estimated level), then the process of learning from mistakes
L
is activated which
modifies the sets: the knowledge of the agent
M
and the set of strategies
•
S
respectively.
The process of learning from its ownmistakes is illustrated in a schema inFig.
3.10
.
The process of learning through imitation is based on the fact that the agent may
observe the behaviour of other agents and particularly the changes which result
from their activity (the realization of their strategies) in the environment (Fig.
3.11
).
The process of learning through imitation takes place when a given agent observes
the surrounding environment and the events that happen in this environment. This
process may be realized in the following stages:
agent
I
V
m
q
s
X
L
the
procedure
of learning
from
mistakes
m
m
I
V'
Fig. 3.10
Schema illustrating the process of the agent's learning based on mistakes
