Malvrok
Reputation: 389
Backpropagation in Pooling Layer (Subsamplig layer) in CNN
My doubt is how do I backpropagate error in the Pooling layer, because when I calculate the derivative, there is only 1 element of 4 (for example, when using a 2x2 pooling kernel) that affects the result of the feedforward.
Upvotes: 13
Views: 5850
Answers (1)
Ash
Reputation: 4718
Suppose you have a matrix M of four elements
a b
c d
and maxpool(M) returns d. Then, the maxpool function really only depends on d. So, the derivative of maxpool relative to d is 1, and its derivative relative to a,b,c is zero. So you backpropagate 1 to the unit corresponding to d, and you backpropagate zero for the other units.
Upvotes: 14
Related Questions
- The order of pooling and normalization layer in convnet
- Convolutional neural network, how the second conv layer works on the first pooling layer
- Exchange a pooling layer using conv2d layer in keras
- Backpropagation for Max-Pooling Layers: Multiple Maximum Values
- Pooling Layer vs. Using Padding in Convolutional Layers
- Max pooling layer after 1D convolution layer
- Convolution and Pooling in TensorFlow Deep MNIST
- What is the advantage of not having pooling layers in between convolutions in this context?
- Backpropagation in Convolutional Neural Networks
- could not calculate the dimensions after pooling and subsampling in the famous Convolutional neural nets example