Pooling Layer¶
- A introduction pooling layer is in here (http://cs231n.github.io/convolutional-networks/#pool)
- In Resnet, we only need to implement a global average pool: which downsample each channel into one value([N,C,H,W] -> [N,C,1,1]). Take a looked at a reference global average pool implementation in numpy. (it’s private repo but i have sent requests to add you as collaborators.). Also takes a look at the backward pass.