Question 39
UnclassifiedWhat is the primary purpose of max pooling layers in a convolutional neural network (CNN)?
Correct answer: B
Explanation
Max pooling layers reduce the spatial size of feature maps by taking the "maximum activation in each region," which downsampled the data while preserving the strongest signal. This also gives the network some "translation invariance" because small shifts in the input are less likely to change the pooled output.
Why each option is right or wrong
A. Increase the spatial resolution of feature maps by interpolating new pixels between existing activations.
B. Downsample feature maps by selecting the maximum activation in each region, providing some translation invariance.
Max pooling is the standard CNN operation that partitions each feature map into fixed-size windows and outputs the largest value from each window, so the representation is compressed while the strongest detected feature is retained. Because the exact location of that peak can move slightly within the pooling window without changing the output, the layer makes the model less sensitive to small translations in the input.
C. Add non-linearity to the network by applying an activation function such as ReLU.
D. Train the convolutional filters by updating their weights during backpropagation.