Lesson 16 Pooling
Pooling is one of methods to reduce image size, keeping the characteristics, like convolution. This algorithm is valid to reduce the number of inputs for neural network.
The process is similar to convolution, but the filter has no parameter. The system extracts a part of an image with the filter, and get a value after a calculation on the pixel values in it. The process to get a maximum value is called Max Pooling, and to get an average value Average Pooling.
Now, we are going to see how it works with a grey scale image. We suppose the same grey scale image as the convolution process case, and conduct max pooling with a filter of three pixels in width and height.
As a result, we get an image with three pixels in width and height.
Pooling process creates a new image with smaller pixel size, keeping the characteristics of an original image, as convolution process. This process allows us to reduce the total time for learning process, because of less number of parameters for learning.