Convolutional network (CNN)

Network的架构调整

1、All the images to be classified have the same size.

Receptive filed

Simplification 1 - Typical Setting

all channels : 会看所有的channels

kernel size : 长和宽 （e.g., 3*3)

Stride : 移动的步长,希望有高度的重叠

padding : 补值，补充超出范围的值

Each receptive field has a set of neurons.

Each receptive field has the neurons with the same set of parameters, which called filter.

Pooling

Flatten