用分组神经网络识别手写体数字

Handwritten Digits Recognition with Subgroup Neural Network

用神经网络识别手写体数字,大多数采用的是单个的神经网络结构.本文给出一个基于特征输入的手写体数字识别的分组神经网络结构,它是一个混合网络(Hybrid System).实验的结果表明,同只采用单个神经网络结构的方法相比,分组网络具有更好的识别结果.

Most of the neural network handwritten digits recognition systems adopt a net- work with single structure.In this paper,we implement a subgroup neural network structure based on feature description to recognize handwritten digits.It is noticeable that this sub- group network is a hybrid system.The experiment result indicates that the subgroup network gets much better performance than usual single network structure.