Dynamic learning in genetic programming

Generic programming (GP) has been shown useful in constructing non-linear models by learning the sample data The appropriate use of genetic operators usually is believed effective to improve the learning results. This paper describes a dynamic mutation strategy to expedite the evolution process. Suitable mutation rate along with its corresponding mutation operator is dynamically selected and proceeded according to the resulting fitness values. The introduction of mixed genetic operators extends GP in adapting better generation of program solutions. The comparable learning performance of different mutation operators applied to various data sets are discussed.