Learning by adapting representations in genetic programming

Machine learning aims towards the acquisition of knowledge, based either on experience from the interaction with the external environment or by analyzing the internal problem-solving traces. Genetic programming (GP) has been effective in learning via interaction, but so far there have not been any significant tests to show that GP can take advantage of its own search traces. This paper demonstrates how an analysis of the evolution trace enables the genetic search to discover useful genetic material and to use it in order to accelerate the search process. The key idea is that of genetic material discovery which enables a restructuring of the search space so that solutions can be much more easily found