The keras package contains the function application_vgg16 that loads the VGG16 model for image classification. Load this model into R and print out the model. In each layer, the number of trainable...

The keras package contains the function application_vgg16 that loads the VGG16 model for image classification. Load this model into R and print out the model. In each layer, the number of trainable weights is listed. What proportion of trainable weights is in the convolutional layers? Why is this such a small portion of the entire model? In other words, why do dense layers have many more weights than convolutional layers?