What is pooling in Convolutional Neural Network? | What is Flattering in Convolutional Neural Network?

Pooling is one of the other building blocks of a Convolutional Neural Network.
It is a non-linear layer and the key role is to downsample or reduces the dimension. It is useful in reducing the variance, amount of parameters, and computation in the network, it is helpful in extracting sharp and smooth features of the input image.

There are two types of pooling:

•  Max Pooling – it extracts the low-level features like dot or point, edge, etc, and used vastly.

How do we perform Max Pooling?

For 1-Dimensional pooling, we take a stride of 2 and pool the maximum from them.

 

 For 2-Dimensional pooling, we take stride of 2x2 and pool the maximum from them.

• Global Average pooling – high-level smooth features.

How do we perform Global Average Pooling?

Addition of all the elements and then dividing by total numbers of rows and columns, and it is always unit.

What is Flattering?
Flattening is converting the multi dimensional data array into a 1-dimensional array from output of a previous layer create a single long feature, and it is connected to the final fully-connected layer in Convolutional Neural Network.