Performance improvement of machine learning via automatic discovery of facilitating functions as applied to a problem of symbolic system identification

The recently developed genetic programming paradigm provides a way to genetically breed a population of computer programs to solve problems. The technique of automatic function definition enables genetic programming to define potentially useful functions dynamically during a run, much as a human programmer writing a computer program creates subroutines to perform certain groups of steps which must be performed in more than one place in the main program. An approximation is found to the impulse response function, in symbolic form, for a linear time-invariant system. The value of automatic function definition in enabling genetic programming to accelerate the solution to this illustrative problem is demonstrated