Machine Learning Based on Coevolutionary Genetic Algorithms
Pei Jiang · Hu'nan Shifan Daxue xuebao. Ziran kexue ban · 2004
Presented a method of machine learning based on coevolutionary genetic algorithms. The method coevolves the training environment with respect to the population of learners so that the training environment provides an evolutionary gradient for learners, and it overcomes the difficulty of the performance of a learning system relying heavily on pre-knowledge of problems. The method is applied to a rule learning task of one-dimensional cellular automata, with the experiment result showing an improvement for the problem compared to the best known result.