Exercises
These exercises check the convolution ideas from the chapter.
A local patch is 5 x 5. How many positions does it contain?
Compute it first, then check your number.
Input window [3, 1, 2] uses kernel [2, 0, -1]. What is the output?
Compute it first, then check your number.
A convolutional layer has 20 kernels. How many feature maps does it produce?
Compute it first, then check your number.
For input length 10, kernel size 4, and stride 2, what is floor((10 - 4) / 2) + 1?
Compute it first, then check your number.
A kernel is 5 x 5 and reads 3 input channels. How many weights does it have?
Compute it first, then check your number.
What is the max of [7, 2, 5, 9]?
Compute it first, then check your number.
Enter 1 if convolution reuses the same kernel across positions, or 2 if it always learns a separate kernel for every position.
Compute it first, then check your number.