We introduce a new form of linear genetic programming (GP). Two methods of
acceleration of our GP approach are discussed: 1) an efficient algorithm th
at eliminates intron code and 2) a demetic approach to virtually paralleliz
e the system on a single processor. Acceleration of runtime is especially i
mportant when operating with complex data sets, because they are occuring i
n real-world applications. We compare GP performance on medical classificat
ion problems from a benchmark database with results obtained by neural netw
orks. Our results show that GP performs comparable in classification and ge
neralization.