一种在非单调和不完全领域知识下的EBL方法

An EBL Method under Nonmonotonic and Incomplete Domain Knowledge

基于解释的学习（ＥＢＬ）克服了基于相似性的学习中的一些局限性，但是由于领域知识的不完全性和非单调性，在ＥＢＬ过程中会出现多重解释问题，运用分层ＡＴＭＳ（基于解释的真值维护系统）来实现ＥＢＬ方法．由于分层ＡＴＭＳ能处理非单调推理和多重假设集，因而它能根据不同原因解决ＥＢＬ中的多重解释问题．

Explanation-based Learning can overcome some limitations in similarity-based learning. But due to the nonmonotonic or incomplete domain theories, multiple explanation problems may arise in EBL. This paper presents an approach by using stratified ATMS to realize EBL. Because stratified ATMS can deal with nonmonotonic reasoning and multiple assumption sets, it can solve multiple explanation problems in EBL according to different reasons.