权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Knowledge-Base including Advanced Intelligent Function by handing Inconplete Knowledge

通过处理不完整的知识，包含高级智能功能的知识库

基本信息

批准号：
02452154
负责人：
ISHIZUKA Mitsuru
金额：
$ 3.58万
依托单位：
University of Tokyo
依托单位国家：
日本
项目类别：
Grant-in-Aid for General Scientific Research (B)
财政年份：
1990
资助国家：
日本
起止时间：
1990 至 1991
项目状态：
已结题

项目摘要

For expanding the capability of current knowledge-base systems by adding advanced artificial intelligence functions onto deductive inference mechanism, we carried out our research from the viewpoint of handling incomplete knowledge. Incomplete knowledge here means the knowledge with exceptions, hypothetical knowledge, defeasible knowledge, etc. We have selected a hypothetical reasoning system as our framework, since it can handle incomplete knowledge as hypothesis and is practically important framework applicable to many problems. The crucial problem with the hypothetical reasoning system is its slow inference speed because of its non-monotonicity nature. Thus we focused our research work on finding fast inference mechanism for the hypothetical reasoning.One achievement is a fast hypothetical reasoning method using inference-path network which contributes to avoid backtracking due to the inconsistency among hypothesis. Furthermore, in order to overcome the worst-case limit of exponential-order inference time, we have constructed several efficient hypothetical reasoning mechanisms. Namely, they are a hypothetical reasoning method using analogy, an experience-based learning mechanism for efficient hypothetical reasoning, logical knowledge compilation method foe efficient abductive hypothesis synthesis and a hypothetical reasoning method based on 0-1 integer programming with approximation.

为了扩展现有知识库系统的能力，在演绎推理机制中加入先进的人工智能功能，我们从处理不完整知识的角度进行了研究。不完全知识在这里是指有例外的知识，假设知识，可废止的知识，等等。我们选择了一个假设推理系统作为我们的框架，因为它可以处理不完全的知识作为假设，是实际上重要的框架适用于许多问题。假言推理系统的关键问题是其推理速度慢，因为它的非单调性。因此，本文的研究工作主要集中在为假设推理寻找快速推理机制上，其中一个成果是利用推理路径网络实现了一种快速假设推理方法，避免了假设间不一致性导致的回溯。此外，为了克服指数阶推理时间的最坏情况限制，我们构造了几种有效的假设推理机制。即利用类比的假设推理方法、基于经验的高效假设推理学习机制、基于逻辑知识的高效溯因假设综合方法和基于0-1整数规划的近似假设推理方法。