Databases Reference
In-Depth Information
the problem at hand. If pheromone decays too quickly then good solutions
will lose their appeal before they can be exploited. If the pheromone decays
too slowly, then bad solutions will remain in the system as viable options.
Randomness:
The primary driving factor in this example is
randomness. Where piles start and how they end is entirely determined
by chance. Small fluctuations in the behavior of termites may have a
large influence in future events. Randomness is exploited to allow for new
solutions to arise, or to direct current solutions as they evolve to fit the
environment.
Multiple Interactions:
It is essential that many individuals work
together at this task. If not enough termites exist, then the pheromone
would decay before any more pebbles could be added to a pile. Termites
would continue their random walk, without forming any significant piles.
Stigmergy:
Stigmergy refers to indirect communications between
individuals, generally through their environment. Termites are directed to
the largest hill by the pheromone gradient. There is no need for termites
to directly communicate witheachotheroreventoknowofeachothers
existence. For this reason, termites are allowed to act independently of
other individuals, which greatly simplifies the necessary rules.
Considering the application of intelligent agents segregated in different
chapters of this topic one should also expect much more applications in
various domain. We do believe that the method based on intelligent agents
hold a promise in application to knowledge mining, because this approach
is not just a specific computational tool but also a concept and a pattern
of thinking.
1.4. Summary
Let us conclude with some remarks on the character of these techniques
based on intelligent agents. As for the mining of data for knowledge the
following should be mentioned. All techniques are directly applicable to
machine learning tasks in general, and to knowledge mining problems in
particular. These techniques can be compared according to three criteria:
e
ciency, effectivity and interpretability. As for e
ciency, all the agent
based techniques (considered in this chapter) may require long run times,
ranging from a couple of minutes to a few hours. This however is not
necessarily a problem. Namely, the long running times are needed to find
a solution to a knowledge mining problem, but once a solution is detected,
