Pooling is a procedure that takes input over a certain area and reduces that to a single value (subsampling). In convolutional neural networks, this concentration of information has the useful property that outgoing connections usually receive similar information (the information is “funneled” into the right place for the input feature map of the next convolutional layer). This provides basic invariance to rotations and translations. For example, if the face on an image patch is not in the center of the image but slightly translated, it should still work fine because the information is funneled into the right place by the pooling operation so that the convolutional filters can detect the face.