$25
Homework-9
Question 1
Consider the following deep learning model, with layers stacked in a sequential order:
layer 0: RGB image of size 200 × 200
layer 1: MaxPooling2D layer with pool size (2, 2)
layer 2: Conv2D layer with 64 filters and kernel size (3, 3). The padding method is “same”. .
layer 3: MaxPooling2D layer with pool size (2, 2)
layer 4: Flatten layer
layer 5: Dense layer of 10 nodes
layer 6: Dense layer of 4 nodes
layer 7: SoftMax
1
Nonlinear activation (ReLU) should be applied immediately after some of these layers. Mark below the
layers that should be followed by ReLU.
Answer
0 / 1 / 2 / 3 / 4 / 5 / 6 / 7
2
Compute the number of trainable weights that are associated with each layer. (These are the weights with
values that are determined during the training.) For simplicity assume no bias connections. Show your
calculations.
Answer
layer 0: weights.
layer 1: weights.
layer 2: weights.
layer 3: weights.
layer 4: weights.
layer 5: weights.
layer 6: weights.
layer 7: weights.