What is adaptive pooling and why is it useful?

easy

Answer

$(H_{\mathrm{out}}, \;W_{\mathrm{out}})$
$Standard trick to accept variable input resolutions with a fixed-size classifier head. torch.nn.AdaptiveAvgPool2d((1,1)) is the go-to for global pooling.$
$Also useful for detection networks that need fixed-size feature maps for RoI heads.$

Check yourself — multiple choice

Only works on 1x1 outputs
Automatically picks kernel/stride to yield a target output size → supports variable input resolutions
Same as adaptive learning rate
Removes all weights

Adaptive pool: auto kernel/stride for any input → fixed-size output → supports variable inputs.

#cnn

214 interview questions in this topic.