一类递归逻辑程序的强构造学习

STRONGLY CONSTRUCTIVE LEARNING OF A SORT OF RECURSIVE LOGIC PROGRAM

定义了一类强构造学习问题 ERP;探讨了以数据驱动方式提高强构造学习系统效率的途径 ,特别是学习系统的主动采样方式与系统性能的关系 ;给出了一个效率较高、输入条件基本实用的 ERP(1)学习算法 ,该算法原则上可推广到任意 n>1的 ERP(n)问题 .

In this paper, a sort of strongly constructive learning problem——ERP is defined; an approach to design efficient ERP learning algorithms is introduced; and the relation between active sampling and efficiency is clarified. Moreover, a relatively efficient ERP (1) algorithm is provided, which imposes a weakest extra demand on the input and can be generalized to any ERP(n ) for n >1.