10

Assume that an RGB image input enters a ConvNet: 0 4 6 4 5 4 7 3 1 1 2 8 4 70 20 4 2 5 5 2 2 8 1 7 2 1 1 6 1 2 3 2 8 4 70 3 8 0 3 5 4 2 3 3 1 2 7 3 1 1 2 7 1 4 2 2 7 2 7 0 43 3 5 1 6 0 2 3 2 For the convolution layer, a single vertical edge filter (shown below) with the padding size of 2 and the stride of 2 with ReLU activation and no bias was used. 1 0 -1 1 0 - 1 1 1 0 -1 1 0 -1 1 0-1 1 0 -1 10 -1 1 0 For the pooling laver, max pooling with with =2 and f=2 was used. The output of the pooling layer flatens and enters the fully connected network The output of the pooling layer is 2x2 size. (Xj j represents the element in the output of the pooling layer in the i-th row and j-th column) X1,1 = X1,2 = X2 1 = X2.2 =