well if max pooling layers only extract the valuable features by sliding over the image and it does not have any parameters unlike convolution layer to be leant while training .It only have hyperparameters like size of the filter to be chosen then it means that pooling layer doesnt play any role while backpropogation(when we are learning paramters)???
Regarding Pooling Layers
hey @amankharb ,
Yes, you are absolutely correct that MaxPoolingLayers doesn’t play any role in backpropagation .
But it does something , which helps in backpropagating more appropriately.
As , what is helps is in , it reduces computations by taking only the maximum values , which means
-> more fast computation
-> less computation power required.
-> Saves memory while allocating values.
-> Somewhat help in faster convergence of model.
I hope this would have resolved your doubt and helped in understanding the core use of MaxPooling Layer in Back Propagation.
Thank You and Happy Learning 
.
so it only speeds up the the calculation of derivatives while backpropogation and it does not effect the value of derivatives?
Yes You are right. It just lowers the count of variables , by just the maximum or average value, which means the value may or may not change , but at the number of values is decreased so calculating derivatives becomes much faster.
I hope this helps.
Thank You.
I hope I’ve cleared your doubt. I ask you to please rate your experience here
Your feedback is very important. It helps us improve our platform and hence provide you
the learning experience you deserve.
On the off chance, you still have some questions or not find the answers satisfactory, you may reopen
the doubt.