Pattern recognition generally aims to classify or fit a function to a set
of observations and is applied by several CSSIP Programs. Most programs
have points of collaboration with this Pattern recognition, with scope
for more. A significant focus is on adaptive pattern recognition, where
the system makes decisions about the form of preprocessing and the class
of experts to use before classifying or function fitting. Artificial
neural network and statistical methods are proving to be very effective
in a number of CSSIP tasks. As with any pattern processing method,
the quality of the information provided as input is vital and therefore
we are closely involved with preprocessing issues, as it is not possible
to divorce the preprocessing issues from the more classical decision process.
For many tasks, human cognitive processes are very effective. Such processes
depend on biological neural nets, thus providing stimulus to study neural
net principles. A neural net approach sometimes suggests a solution to
a specific task which reveals a direct algorithmic method. Recent
efforts in adaptive pattern recognition include studies of networks that
make decisions about the form of preprocessing, and the class of experts
to use before classifying or function fitting. The study of geometric
mapping of concepts is a closely related work, which is relevant to the
presentation of data to facilitate human decisions on patterns. An overarching
viewpoint is that systems for decision making should make optimum use of
the combination of automatic adaptive methods and human perception and
cognition. Pattern recognition generally aims to classify or fit a function
to a set of observations and is applied by several CSSIP Programs.
Closely involved with preprocessing issues, as it is not possible to divorce
the preprocessing issues from the more classical decision process.
