Growth Phenomenon of Individuals' Code Length in Genetic Inductive Logic Programming
Yang Xin · Journal of Computer Research and Development · 2003
The individuals' code growth existing in GILP badly reduces the algorithm performance and the readability of rules. Through the schema analysis of individuals with variable length, the individuals' code growth existing in GILP is explained. In addition, the population will degenerate when the length penalty of individual's fitness is added from the initial population. If the new penalty strategy is employed and the length penalty of individual's fitness is not added until the population properties are obviously improved, then the population's degeneration is avoided and the individual's code growth is also effectively restrained.