解释与智能系统  

Explanation and Intelligence Systems

作　　者：杨莉[1] 葛建新[1] 

机构地区：[1]浙江大学计算机系,杭州310027

出　　处：《计算机科学》1994年第2期6-9,共4页Computer Science

基　　金：国家自然科学基金;浙江省自然科学基金

摘　　要：1,前言解释在人工智能(AI)系统中是一个关键的功能。在基于事例的推理中.当一个预期的断言失败时.解释用于调整知识的结构,即基于错误驱动的学习,在某些情况下,用户不是领域专家,但用户有接受或拒绝由AI系统所产生结果的权力。7n this paper,a brief overview of explanation in symbolic artificial intelligence is given firstly,and then various proposals for explanation in feedforward perceptrons and relaxation-type neural networks (NN) are discussed, and several successful implementations of explanation components in NN systems are presented. Tt is shown that the introduction of structure mechanism (e. g. the explicit encoding of relations and modular network architectures) to a network significantly facilitates the generation of explanation structures in NN.

关 键 词：人工智能 解释 智能系统 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]