Information Technology Reference
In-Depth Information
respect to count values is probably a good one, because it is statistically consistent
and also consistent with the real world. Ideas from nature or the physical world
support this and are interesting, but should probably not be taken too seriously. They
could introduce a very light form of intelligence, although a light form of knowledge
is required
ned by an
automatic count. The rule set can then give additional structure independently, but it
is still the presented data that determines what trees get built. Problems with the
process might include the creation of a long list of very short trees that represent
nothing in particular by themselves. This then begins to look a bit like the standard
memory storage on a computer, with pointers between pieces of memory linking
them up. There is however the possibility of building larger more meaningful trees as
well. A comparison, or relation, with Markov models has been introduced because
they are known to work well and may exhibit the same statistical counting property.
The construction process builds hierarchies automatically and these can repre-
sent any type of concept. A slightly weaker idea is therefore to try to build service-
based business processes or compositions in the same way, where the earlier
stigmergic links were suggested for the same task. See, for example (Greer
2008
,
Sect. 7.3.2.1), or maybe (Aslam et al.
2007
) or (Atkinson et al.
2007
). While real-
world concepts or natural language might be restricted by sets of relations that can
justify the triangular counting mechanism, more complex business processes might
not be. There is a difference between a sub-process and linking two independent
processes. In that case, statistical counts would be used purely for reliability, but it
is a known problem and several solutions that are at least semi-automatic, have
already been suggested. It is worth noting that the count values could be used as
probability values, or something similar, as each tree is a bit self-contained. If a
particular structure was presented to a network and one of the concepts was
missing, the system could try to calculate a probability value, indicating the con-
first. Any concept is allowed to be a main one and this is de
fidence that the missing concept was in fact an error. This would be an automatic
way to assign a value range to the stored data, for security reasons, or other. So
concept trees can also be looked at in terms of automatically creating process
hierarchies and really does span from the large Internet-based network to the small
cognitive model.
Not every group of concepts should be added either and dynamic factors like
reinforcement and time can also be considered. So while the construction process is
automatic, a reasoning component might also make certain decisions. For example,
does a link to another newly created tree actually apply to my instance? If a real tree
is taken as the natural world model, and why not, then it obeys the rule that a
heavier branch will cause a lighter one to snap. The new AI part then is the idea of
an intelligent indexing and linking system, to keep consistency between the split
trees. This means that even if the original structures disintegrate, while the natural
world entity would tend to chaos, the linked elements will allow for traversal
through speci
c channels and maintain the order. The question would be how
ef
cient or accurate the structure can be.
Search WWH ::
Custom Search