(1) concept
Pooling layer, layer to relieve excessive sensitivity convolution position.
With convolutional layer, the shape of a fixed pool of input data window layer (pooling window) to calculate the output element, from the upper left, from left to right, top to bottom, the input data are sequentially slides . Pooling computing a maximum value or an average value layer pooled window elements, we are called mean maximum pool and pooling.
Below, the maximum pooled,
(2) filling and stride
Like convolution layer, pooling layer may also change the shape of the output of the input moving stride aspect sides or resize the window, and filled
(3) Multi-channel
When a multi-channel input data, the number of cell layer is equal to the number of input channels and output channels for each input channel are pooled, the pooled layer