【摘要】 知识可增殖人工神经网络是一个由多个已知功能人工神经网络和匹配人工神经网络协同工作共同组成的具有扩展功能、知识继承和积累作用的大规模人工神经网络。文章使用微分几何作为数学工具来分析、描述神经网络的场理论,研究可学习的机构和已知功能人工神经网络之间的组织结构协同工作方法。并利用统计流形整体不变几何结构特性为理论基础,完成层次化混合专家神经网络的信念分配,实现知识可不断扩充,功能不断增强的人工神经网络。更多还原

【Abstract】 NN is a kind of large-scale artificial neural network with the enlargement function.It composes of several known ar-tificial neural networks and matches the NN to cooperating to-gether.We make use of differential geometry as mathematical tool to analyze and describe the field theory of NN as well as study realization of the cooperating together from the point of view of the NN field theory.We achieve the distribution of HMNN and Knowledge-increasable ANN based on the theory of manifold global invariant properties.更多还原