Information Technology Reference
In-Depth Information
but blindly presumed that input presented at the same time must simply be related.
The original cognitive model did include an ontology or knowledge-base reference,
to provide this type of support. Some comparisons with bio-related models in
general can also be made.
7.1 Biological Comparisons
More recent work again (Greer 2013a ) has put Hebb
'
s well-known theory of
'
neurons that wire together,
re together
'
, into a computer model. It has added a
mechanism using the idea that when they
re together they may be attracted to each
other and grow links to join up. The rules of Sect. 5.4 (maybe point 5.b)
fit in well
with this as it suggests comparing the sets of input links to trees. If both trees have
the same set of input links, then when these
fire, both trees will be activated and can
therefore decide to join up. It does not however suggest exactly how they might
grow towards each other or combine in a biological model. The earlier cognitive
model (Greer 2008 , 2013b )de
nes a 3-layer architecture, where the bottom level
links for optimisation purposes, the middle layer links to aggregate these pattern
groups, while the top layer links to create higher-level concepts and trigger dynamic
events. As the concept trees theory is more about aggregating and balance, over all
of the data, it is more suited to this middle level. It has also been noted in the formal
speci
cation of Sect. 5 that a concept might be duplicated, simply because of the
distributed nature of the system. This is also the case for the human brain, as it is
known to duplicate information and the practical aspects of trying to access a
particular brain region might make it easier to simply duplicate some information
locally. Ideas of entropy and automatic monitoring can also be related to both the
stigmergic/dynamic linking model (Greer 2008 , Sect. 8.5) or the concept trees. As
either system develops, it will tend towards some sort of
fixed structure. This trend
would then only be broken by a change in the input state. So, after the formations
are created, any more dramatic changes might indicate a change in data, and so on.
This idea probably applies to most dynamic systems with similar designs.
7.2 Higher-Level Concepts
There is a reference to a linking structure in Greer ( 2008 , Sect. 9.3.7, Fig. 24) that
describes how linked concepts might only be related or activated if they are
assigned speci
c values. For example, if we have mother, son and uncle concepts
linked, then it might only be true if the mother is called Susan, the son is called John
and the uncle is called David. The idea of pushing the descriptive text to the leaf
nodes, so as to represent speci
c instances, has been written about in Sect. 6.3 .
There is also a reference to another linking structure in Greer ( 2008 , Sect. 9.3.7,
Fig. 25) that tries to index different concept sets through unique keys. It has the
Search WWH ::




Custom Search