- Size of the region in the input that produces a pixel value in the activation map

- The larger the receptive field, the more information the activation map contains about the entire image

- We can increase the receptive fields by adding more layers

- This requires less parameters than increasing kernel size

- Pooling:

    - increases the receptive field 

    - reduces the number of parameters

    - preserves important features

    - makes images immutable to smaller changes

- Retains the information within a specified region 

- Collects spatial variants that helps the neural network to identify patterns in images

- Reduce the dimension by pooling out the sumarized information

- Max pooling : Pool out the max value in the specified region

- Average pooling : Pool out the average value in the specified region