RESEARCH ON A GENETIC PROGRAMMING PARADIGM USING LINEAR REPRESENTATION

One of important subjects in the field of genetic programming （GP） is how to represent the evolutionary individuals （computer programs） efficiently. In this paper, the individuals with linear postfix forms of tree structures are encoded using position information to perform various genetic operators. The related formal definitions are given. A stack-based GP algorithm is designed and realized. The performance of the operations is examined using simulated experiments. The encoding manner can be extended to the linear structures of programs to implement specialized operators, which shows that the method possesses the feasibility and flexibility in dealing with different problems. The schema theory for linear GP with one-point crossover of string-based is also proposed, which may unify the schema creation mechanism of the standard GA into this theoretical frame.