2. Suppose that a convolutional layer of a neural network has an input tensor X and computes an output via correlation and ReLU activation, i.e., Z[i, m, j1, j2] =W[m, n, k, k2]X[i, n, j1 + k1, j2 +k2] +b[m], n k1 k2 U[i, m, j1, j2] = max {0, Z[i, m, j1, j2]}. for some weight tensor W and bias vector b. Suppose that X has shape (1,10,48,64) and W has shape (20,10,3,3). Assume the convolution is computed on the valid pixels. (a) What are the numbers of input and output channels, and how big are the convolution kernels? (b) What are the shapes of Z and U? (c) How many multiplications must be performed to compute Z? (d) If W and b are to be learned, what are the total number of learned entries?