跳转至

05

Convolution

We can compute the spatial size of the output volume as a function of the input volume size (\(W\)), the receptive field size of the Conv Layer neurons (\(F\)), the stride with which they are applied (\(S\)), and the amount of zero padding used (\(P\)) on the border. You can convince yourself that the correct formula for calculating how many neurons “fit” is given by \((W−F+2P)/S+1\).

Pooling

  • Accepts a volume of size \(W_1 \times H_1 \times D_1\)

  • Requires two hyperparameters:

  • their spatial extent \(F\),

  • the stride \(S\),

  • Produces a volume of size $W_2 \times H_2 \times D_2 $

where:

  • \(W_2 = (W_1 - F)/S + 1\)
  • \(H_2 = (H_1 - F)/S + 1\)
  • \(D_2 = D_1\)

Normalization

Fully-Connected

经典框架

LeNet:90 年代的卷积神经网络

AlexNet:更深的卷积神经网络

GoogLeNet:采用了 Inception Module

VGG:深度达到了 19 层,但训练的稳定性不是很好

ResNet:通过 skip connection 和残差模块学习残差,网络更深,分类效果更好

卷积神经网络可视化

权值可视化

遮挡

梯度上升

迁移学习

新数据集很小,但和原数据集相似度高

训练全连接分类器

新数据集很大,和原数据集相似度高

可以考虑训练整个网络

新数据集很小,并且和原数据集相似度低

还是单独训练一个 SVM 之类的吧

新数据集很大,但和原数据集相似度低

完全可以从头开始训练,但是从预训练模型开始训练是比较好的选择