Learning theory for higher-knowledge self-organization from experiences

本文的目的是建立从经验中获取一般知识的学习理论和算法。所提出的方法具有一个层次结构，由一组第一学习者和第二学习者，其中第一学习者估计模型的个人经验和第二学习者倾斜的第一学习者的学习结果。 我们发现，该任务被描述为张量方程，它可以解决的自组织映射（SOM）的高阶。在此基础上，我们提出了基于高阶SOM的张量SOM。

The purpose of this work is to establish the learning theory and the algorithm acquiring the general knowledge through experiences. The proposed method has a hierarchical structure consisting of a set of first learners and the second learner, in which the first learners estimate models of individual experiences and the second learner leans the learning result of the first learners. We found that the task is described as tensor equations, which can be solved by the higher-rank of self-organizing map (SOM). Furthermore, we developed tensor SOM based on the higher-rank of SOM.