Incomplete Information Processing and Self-organization of Deep Knowledge for Intelligent Systems

本研究提出了一种新的建模方法，该方法考虑了人类有选择地提取符合其观点的信息并自适应地组织这些信息以达到解决问题的目的的能力。并开发了基于该模型的问题求解过程的计算机算法。主要研究成果如下：提出了一种提取知识表示单元对应的原始事件的方法。具体的因果序列。也就是说，‘正弦’是在它们的序列中形成的。给出了基于表示元组织具体因果序列所涉及问题的深层结构的推理方法，阐述了从现实世界中不完整、有时不一致的信息中产生完整一致知识的过程，该过程基于统治世界的秩序关系，例如，在公寓里，禁止养狗是一种默契，如果情况是禁止养猫，则是真的。提出了一种将来自环境的信号或符号等原始信息转化为符号表示的知识的方法，该方法基于表示所考虑问题域的约束的区间约束，引入了约束区间模糊集的概念。同时，将演绎分组、基于解释的分组与神经网络、遗传算法等数据分组方法相结合，提出了一种基于分布式感知处理单元的分布式协作问题求解方案，从局部信息生成全局一致的系统状态描述.还有。同样的问题现在继续被表述为‘语法学’问题，这是一种在生物中创造秩序的理论。

In this study, a new modeling method is proposed, in which the human abilities to extract information selectively corresponding to his or her point of view and to organize it adaptively for the purpose of problem solving is treated. And the computer algorithms executing the problem solving processes based on the model are developed. The main results obtained are as follows :A method is presented which extracts the primitive events corresponding to knowledge representation units. Concrete cause-effect sequences. that is, ' sinarios ' are formed in their sequences. Also, the inference method to organize the deep structures of problems involved in the concrete cause-effect sequences is presented based on the representation units.The process to create the complete and consistent knowledge from the incomplete and sometimes inconsistent information from the actual world is formulated, which is based on ' order relations ' goveming the world for instance, in an apartment, Keeping dogs are prohibited is a tacit understanding, if the case of, keeping cats are prohibited is true. ').A method to transform raw information such as signals or signs from the environment into the knowledge represented as symbols is proposed, in which the notion of ' constraint interval fuzzy sets ' is introduced based on ' interval constraints ' which represent the constraints goveming the concemed problem domain. Also, the same problem is treated with the method combining the deductive Teaming EBL (Explanation-based teaming and the data teaming methods such as the neural network and the genetic algorithm techniques.A method to create the global and consistent descriptions of the system states from the partial information is presented based on the decentralized cooperative problem solving scheme by distributed sensing and processing units. Also. the same problem is now continuing to formulate as the problem of ' synagetics which ia a theory of creating order in a living thing.