Given an input array X X X and a kernel/filter K K K as follows:
X = [ − 1 − 1 0 2 − 2 1 0 0 1 − 1 − 1 0 0 0 0 1 ] X = \begin{bmatrix} -1 & -1 & 0 & 2 \\ -2 & 1 & 0 & 0 \\ 1 & -1 & -1 & 0 \\ 0 & 0 & 0 & 1 \end{bmatrix} X = − 1 − 2 1 0 − 1 1 − 1 0 0 0 − 1 0 2 0 0 1
K = [ 0 1 0 1 1 0 0 1 1 ] K = \begin{bmatrix} 0 & 1 & 0 \\ 1 & 1 & 0 \\ 0 & 1 & 1 \end{bmatrix} K = 0 1 0 1 1 1 0 0 1
Convolve the kernel K K K over the input X X X with a stride s = 1 s = 1 s = 1 and no padding to obtain matrix A A A .
Apply average pooling on A A A to produce matrix B B B .
Pass B B B through the sigmoid (logistic) function to get the final output y ^ \hat{y} y ^ .
Given that ∂ L ∂ y ^ = − 1 \frac{\partial L}{\partial \hat{y}} = -1 ∂ y ^ ∂ L = − 1 , determine the value of ∂ L ∂ K 00 \frac{\partial L}{\partial K_{00}} ∂ K 00 ∂ L , where K 00 K_{00} K 00 is the element of K K K at index ( 0 , 0 ) (0, 0) ( 0 , 0 ) .