AlexNet uses overlapping pooling with
AStride 2 and window size 3
BStride 4 and window size 11
CStride 3 and window size 5
DStride 1 and window size 7
Answer & Solution
Correct answer: A. Stride 2 and window size 3
With s = 2 and z = 3, adjacent 3x3 max-pooling windows overlap. This reduces top-1 and top-5 error by 0.4% and 0.3% and slightly resists overfitting.
Related questions
Image similarity in AlexNet's last hidden layer is measured byThe SGD update rule for AlexNet usesRemoving any single middle convolutional layer of AlexNet costs roughlyAlexNet was trained for roughly how many epochs on ImageNet?At test time dropout is approximated byAlexNet's second form of data augmentation isAlexNet's first form of data augmentation isWhy are ReLU networks faster to train than tanh networks?