Abstract: Characteristic space is a basic concept in artificial intelligence field.Many problems can be simplified or solved by means of a characteristic space transformation method in artificial intelligence field.Here presented is a characteristic space transformation method based on learning from examples.The method can be applied to intelligent control,pattern recognition,knowledge acquisition, etc ..Numerical experiment shows that the method not only possesses a high sense of theory but also possesses a high sense of application.A new method is also given,which improves learning efficiency by means of changing output space description.