Deep Learning Dictionary - Lightweight Crash Course

Max Pooling - Deep Learning Dictionary

Max pooling is an operation that is typically added to convolutional neural networks following individual convolutional layers to reduce the height and width dimensions of image data output from the previous convolutional layer.

When the max pooling operation follows a convolutional layer, the convolutional output from the layer is partitioned into pools where the max pixel value is selected and kept from each pool. All of the non-max values are discarded. The resulting output is an image of reduced dimensions created only with the max pixel values.