神经网络计算机的一种分形实现

A FRACTAL IMPLEMENTATION OF NEUROCOMPUTER

神经网络计算机的实现是神经网络研究领域中的一个重要课题．目前，神经网络的研究已形成了较为系统的理论模型与算法，但神经网络计算机的研究却至今没有重大突破，主要困难就在于网络规模过大，突触联系密度太高等等，为解决这个问题，文中基于分形理论，提出了一种神经网络计算机的分形实现方案，给出了分形维数的计算公式，并从物理上实现了与整体结构具有自相似性的分形子结构，在神经网络计算机的实现上作了有益的探索．

The study of the implementation of neurocomputer is an important research topic in the neural network research field. Several theoretical models and corresponding algorithms have been developed in the theory of neural networks, but there is still little progress in the implementation of neurocomputer, for the problem of large size of network and high density of connections has to be faced. To solve this problem, a scheme of implementation for neurocomputer as well as the formula for fractal dimension are presented in the paper. Physical implementation of a fractal sub structure is given, which is expected to be beneficial for implementing neurocomputer.