卷积神经网络之二：实例及源码示例笔记

来源：互联网发布：java中多线程实现编辑：程序博客网时间：2024/06/05 09:07

文字识别系统LeNet-5

下面，有必要来解释下上面这个用于文字识别的LeNet-5深层卷积网络。

1. 输入图像是32x32的大小，局部滑动窗的大小是5x5的，由于不考虑对图像的边界进行拓展，则滑动窗将有28x28个不同的位置，也就是C1层的大小是28x28。这里设定有6个不同的C1层，每一个C1层内的权值是相同的。

2. S2层是一个下采样层。简单的说，由4个点下采样为1个点，也就是4个数的加权平均。但在LeNet-5系统，下采样层比较复杂，因为这4个加权系数也需要学习得到，这显然增加了模型的复杂度。在斯坦福关于深度学习的教程中，这个过程叫做Pool。

源码：Convolutional Neural Networks (LeNet)

可能用到的源码：logistic_sgd，mlp

结果：

...epoch 180, minibatch 100/100, validation error 0.920000 %training @ iter =  18000epoch 181, minibatch 100/100, validation error 0.920000 %training @ iter =  18100epoch 182, minibatch 100/100, validation error 0.920000 %training @ iter =  18200epoch 183, minibatch 100/100, validation error 0.910000 %     epoch 183, minibatch 100/100, test error of best model 0.920000 %training @ iter =  18300epoch 184, minibatch 100/100, validation error 0.910000 %training @ iter =  18400epoch 185, minibatch 100/100, validation error 0.910000 %training @ iter =  18500epoch 186, minibatch 100/100, validation error 0.910000 %training @ iter =  18600epoch 187, minibatch 100/100, validation error 0.910000 %training @ iter =  18700epoch 188, minibatch 100/100, validation error 0.910000 %training @ iter =  18800epoch 189, minibatch 100/100, validation error 0.910000 %training @ iter =  18900epoch 190, minibatch 100/100, validation error 0.910000 %training @ iter =  19000epoch 191, minibatch 100/100, validation error 0.910000 %training @ iter =  19100epoch 192, minibatch 100/100, validation error 0.910000 %training @ iter =  19200epoch 193, minibatch 100/100, validation error 0.910000 %training @ iter =  19300epoch 194, minibatch 100/100, validation error 0.910000 %training @ iter =  19400epoch 195, minibatch 100/100, validation error 0.910000 %training @ iter =  19500epoch 196, minibatch 100/100, validation error 0.910000 %training @ iter =  19600epoch 197, minibatch 100/100, validation error 0.910000 %training @ iter =  19700epoch 198, minibatch 100/100, validation error 0.910000 %training @ iter =  19800epoch 199, minibatch 100/100, validation error 0.910000 %training @ iter =  19900epoch 200, minibatch 100/100, validation error 0.910000 %Optimization complete.Best validation score of 0.910000 % obtained at iteration 18300, with test performance 0.920000 %The code for file convolutional_mlp.py ran for 527.38m

0 0