Question 3 (20 points): Given an image of size 50*50*3, where 3 represents RGB representation of each pixel 3.1) Suppose in the first fully connected layer of a neural network, there are 1000 neurons, how many trainable variables, sum of the number of weights and bias variables, are in the first layer? For one image and only one fully connected layer of size 1000, calculate the number of multiplication operations in forward propagation 3.2 Suppose you have a convolution layer. For a kernel of size (5,5,3), filter of size 16, stride of 1, and no padding: how many trainable variables, weights and bias, are in the convolution layer? For one image and one convolution layer, calculate the number of multiplication operations in forward propagation