这个是我找到的卷积神经网络在minist库上数字识别最好的结果，准确率99.5 ，比C++版本的更好。程序可以直接编译运行，但是因为要下载两个数据库可能非常慢，需要你修改一下代码跳过去。如果你对cnn很感兴趣，可以找我Q:3617 28654-This is what I found convolutional neural network minist library Digital Identification best results, the accuracy was 99.5 , be

用D-CNN的方法，即CNN卷积层加Fisher Vector做图像分类的MATLAB代码-Use D-CNN approach ( Fisher Vector follows results of CNN convolutional layer as the image feature) to classificate the image

CNN卷积神经网络数字识别MATLAB代码-CNN convolutional neural network digital identification MATLAB code

卷积码是一种有记忆的编码,在任意给定的时间单元处,而且也与前m个输入有关,(Coding, convolutional code is a memory at any given time unit, but also related to m input before,)