| As an evolutionary-based machine-learning method, Genetic Programming (GP) is widely used to solve diverse problems. However, this powerful but still young algorithm has not been tried on program understanding and reproduction for real programs in any general high-level computer language. In fact, such an automatic program understanding and reproduction process is definitely in need not only for software production processes but also for a variety of software engineering activities. This thesis presents such an approach---Program Cloning, which is based on GP and aims to automatically produce general-purpose programs according to an understanding of the target problems. The Program-Cloning method will significantly benefit diverse software engineering topics such as software testing, software quality assurance, safety-critical software designing, software modeling, software reengineering and agile software engineering. Based on our Program-Cloning experiments and the corresponding results, this thesis demonstrates its principle and possible applications. |
