摘要
解释学习是演绎式学习方法，而遗传算法是归纳式学习方法，本文提出的解释学习系统模型ＥＢＬ／ＧＡ，结合两者的优点提高了系统的效用．ＥＢＬ／ＧＡ的特点在于：（１）将宏规则的概念加以扩大，并不局限于可操作的规则．（２）采用结构化遗传算法在庞大宏规则空间内搜索有用的宏规则，以优化宏规则集．（３）采从优胜劣汰的进化思想对宏规则进行有效的管理．这些特点使得ＥＢＬ／ＧＡ在提高系统效用方面比包括ＰＲＯＤＩＧＹ在内的以往工作更具优势．实验表明，ＥＢＬ／ＧＡ模型的学习结果是令人满意的．
Explanation-based learning (EBL) is a deductive learning paradigm, while the genetic algorithm (GA) is an inductive one. In the learning system model EBL/GA proposed in this paper, these two different kinds of learning paradigm are hybridized in order to improve the performance of the system. The learned rules (called macro rules) need not be operational in EBL/GA, thus greatly expands the space of macro rules. GA search effectively in the expanded space for more useful macro rules. The experimental result of EBL/GA model is satisfying. At the end of this paper, two reasons why EBL/GA is better than PRODIGY in utility improvement are given.
出处
《计算机学报》
EI
CSCD
北大核心
1997年第2期125-132,共8页
Chinese Journal of Computers
