Does Maxpool2d have any learnable parameter?

1

I am wondering if maxpool2d in pytorch as any learnable parameter? and if so what is that?
I saw people use self.pool1 = nn.MaxPool2d(2, 2) , self.pool2 = nn.MaxPool2d(2, 2), etc in their models. So i assume there should be some learnable parameters.

The reason that im asking is that im trying to build my own maxpool and wanna make sure im doing it right.

2

Max pooling does not have any learnable parameters.
You can check if with:

pool = nn.MaxPool2d(2)
print(list(pool.parameters()))
> []

The initialization of these layers is probably just for convenience, e.g. if you want easily change the pooling operation without changing your forward method.

1 Like