Simultaneous discovery of detectors and a way of using the detectors via genetic programming

Conventional approaches to problems of pattern recognition and machine learning usually require that the user hand craft detectors for key features in the problem environment. A general approach for simultaneously discovering detectors and a way of combining the detectors to solve a problem via genetic programming with automatic function definition are described. Genetic programming provides a way to genetically breed a computer program to solve a wide variety of problems. Automatic function definition automatically and dynamically enables genetic programming to define potentially useful functions dynamically during a run, and to facilitate a solution of a problem by automatically and dynamically decomposing the problem into simpler subproblems. This approach is illustrated with a problem of letter recognition