Two-Layer-CNN-on-MNIST-master
所属分类:matlab编程
开发工具:matlab
文件大小:42950KB
下载次数:54
上传日期:2018-05-27 08:49:10
上 传 者:
布卡齐
说明: 构建2层卷积神经网络特征提取方法的matlab程序源码
(Constructing matlab program source code for 2-layer convolutional neural network feature extraction method)
文件列表:
cnnConvolve4D.m (1016, 2018-03-25)
cnnPool.m (1538, 2015-06-17)
common (0, 2015-06-17)
common\loadMNISTImages.m (811, 2015-06-17)
common\loadMNISTLabels.m (516, 2015-06-17)
common\t10k-images-idx3-ubyte (7840016, 2015-06-17)
common\t10k-labels-idx1-ubyte (10008, 2015-06-17)
common\train-images-idx3-ubyte (47040016, 2015-06-17)
common\train-labels-idx1-ubyte (60008, 2015-06-17)
loadMNISTImages.m (891, 2018-03-18)
loadMNISTLabels.m (516, 2015-06-17)
myCNN.m (8010, 2018-03-06)
t10k-images-idx3-ubyte (7840016, 2015-06-17)
t10k-labels-idx1-ubyte (10008, 2015-06-17)
train-images-idx3-ubyte (47040016, 2015-06-17)
train-labels-idx1-ubyte (60008, 2015-06-17)
Two-Layer-CNN-on-MNIST-master.rar (20783908, 2018-03-06)
# Two-Layer-CNN-on-MNIST
This is a two-layer convolutional neural network tested on MNIST.
The architecture is images->convolution->pooling->convolution->pooling->softmax, with cross-entropy as its cost function and weight decay.
It can reach an accurancy of 96.34%, of course different random initialization may give different result. This is just a result for one run.
To run the code, just run myCNN.m.
A Chinese version of blog about this implementation can be found at: http://www.cnblogs.com/sunshineatnoon/p/4584427.html
近期下载者:
相关文件:
收藏者: