Introduced on day 36
Pooling
A weight-free operation that summarises a small window (maximum or mean) to shrink a feature map and add local invariance to small shifts.
In detail
Max pooling keeps the strongest response in a window, so the exact position of a pattern within the window no longer matters. Global pooling collapses a whole map to one number per channel and removes the dependence on input resolution. Pooling discards position information, which is a cost if where the pattern occurs matters.