Additional

Question 1

What are the parameters in a convolutional layer?

Answer 1

The parameters in a convolutional layer are the weights of the filters (kernels) and the biases. Each filter has weights corresponding to its size and the number of input channels, plus a single bias term.

Question 2

How do you calculate the number of parameters in a convolutional layer?

Answer 2

Question 3

What does the NHWC format represent in TensorFlow?

Answer 3

NHWC format in TensorFlow stands for:

N: Batch size
H: Height of the image
W: Width of the image
C: Number of channels

Question 4

What does the NCHW format represent in PyTorch?

Answer 4

NCHW format in PyTorch stands for:

N: Batch size
C: Number of channels
H: Height of the image
W: Width of the image

Question 5

How do you calculate the output dimensions of a convolutional layer?

Answer 5

Question 6

What is the shape of an input tensor in TensorFlow with NHWC format for a batch of 16 RGB images of size 64x64?

Answer 6

The shape is
[16,64,64,3].

Question 7

What is the shape of an input tensor in PyTorch with NCHW format for a batch of 16 RGB images of size 64x64?

Answer 7

The shape is
[16,3,64,64].

Question 8

If a convolutional layer has 10 filters of size 5x5 and the input has 3 channels, how many parameters does the layer have?

Answer 8

The number of parameters is:
(5×5×3+1)×10=(75+1)×10=760

Question 9

In a convolution operation, what happens when the filter slides over the input image?

Answer 9

The filter performs a dot product between its weights and the corresponding region of the input image, summing the contributions from all input channels, and adds a bias term to produce a single value in the output feature map.

Question 10

Describe the output tensor shape after applying 8 filters of size 3x3 with a stride of 1 and no padding to a batch of 2 RGB images of size 32x32 in TensorFlow (NHWC format).

Answer 10

The output tensor shape will be
[2,30,30,8], assuming the height and width are reduced by 2 (32 - 3 + 1).

Question 11

Describe the output tensor shape after applying 8 filters of size 3x3 with a stride of 1 and no padding to a batch of 2 RGB images of size 32x32 in PyTorch (NCHW format).

Answer 11

The output tensor shape will be
[2,8,30,30], assuming the height and width are reduced by 2 (32 - 3 + 1).

Question 12

What is the purpose of padding in convolutional layers?

Answer 12

Padding is used to control the spatial dimensions of the output feature map. It adds extra borders to the input image, allowing the filter to fully cover the image borders, and helps maintain the input size.

Question 13

Explain the difference between valid and same padding in convolutional layers.

Answer 13

Valid padding means no padding is added, and the filter only slides within the boundaries of the input image, reducing the spatial dimensions. Same padding adds zero-padding such that the output dimensions are the same as the input dimensions.

Question 14

What is the role of stride in convolutional layers?

Answer 14

Stride controls the step size with which the filter moves over the input image. A stride of 1 means the filter moves one pixel at a time, while a stride greater than 1 means the filter skips pixels, reducing the output dimensions.

Question 15

Calculate the output dimensions for an input of size 64x64 with 3 channels, using a filter of size 5x5, stride 1, and valid padding.

Answer 15

Question 16

n TensorFlow, if you have an input tensor of shape
[32,128,128,3] and you apply a convolutional layer with 16 filters of size 3x3, stride 1, and same padding, what will be the shape of the output tensor?

Answer 16

The shape of the output tensor will be
[32,128,128,16].

Question 17

In PyTorch, if you have an input tensor of shape
[32,3,128,128] and you apply a convolutional layer with 16 filters of size 3x3, stride 1, and same padding, what will be the shape of the output tensor?

Answer 17

The shape of the output tensor will be
[32,16,128,128].

Question 18

What is the formula to calculate the receptive field of a neuron in a convolutional layer?

Answer 18

Question 19

How do batch size and number of filters affect the computational complexity of a convolutional layer?

Answer 19

Increasing the batch size increases the number of images processed simultaneously, affecting memory usage but often improving computational efficiency. Increasing the number of filters increases the number of operations, as each filter produces an additional output channel, directly affecting the computational complexity.

Question 20

If a convolutional layer has an input tensor of shape
[64,3,224,224] in PyTorch and 32 filters of size
7×7 with stride 2 and padding 3, what is the shape of the output tensor?

Answer 20