3. CNNs require a large amount of RAM: Consider a CNN composed of three convolutional layers, each with 3x3 kernels, a stride of 2, and SAME padding. The lowest layer outputs 16 feature maps, the middle one outputs 20, and the top one outputs 30. The input images are RGB images of 28x28 pixels. a. What is the total number of parameters in the CNN? b. If we are using 16-bit floats, at least how much RAM will this network require when making a prediction for a single instance? c. If we are using 16-bit floats, at least how much RAM will this network require when training on a mini-batch of 50 images! 3. CNNs require a large amount of RAM: Consider a CNN composed of three convolutional layers, each with 3x3 kernels, a stride of 2, and SAME padding. The lowest layer outputs 16 feature maps, the middle one outputs 20, and the top one outputs 30. The input images are RGB images of 28x28 pixels. a. What is the total number of parameters in the CNN? b. If we are using 16-bit floats, at least how much RAM will this network require when making a prediction for a single instance? c. If we are using 16-bit floats, at least how much RAM will this network require when training on a mini-batch of 50 images