CNN - Convolutional Neural Networks Convolutional Layer Operations Questions and Answers

Question 1

A convolutional layer in a CNN has an input of size 64x64x16. It uses 32 filters, each of size 3x3, a stride of 1, and 'same' padding. What is the dimensionality of the output feature map?

Accepted Answer

64x64x32

Answer

64x64x16

Answer

62x62x32

Answer

32x32x32

Question 2

Which of the following best describes the primary purpose of a 1x1 convolution operation in a CNN architecture?

Accepted Answer

To act as a channel-wise, fully connected layer for dimensionality reduction or expansion.

Answer

To significantly increase the receptive field of the network.

Answer

To perform spatial down-sampling similar to a pooling layer.

Answer

To detect complex spatial features like edges and corners.

Question 3

In a convolutional layer, what is the effect of increasing the stride value from 1 to 2?

Accepted Answer

It reduces the spatial dimensions of the output feature map.

Answer

It increases the number of parameters in the model.

Answer

It enhances the detection of fine-grained, detailed features.

Answer

It requires the use of 'valid' padding.

Question 4

A data scientist is designing a CNN to classify high-resolution medical images where preserving spatial information in the early layers is critical. Which hyperparameter setting for the initial convolutional layers would be most appropriate?

Accepted Answer

Small filter size (e.g., 3x3), stride of 1, and 'same' padding.

Answer

Large filter size (e.g., 11x11) and a large stride (e.g., 4).

Answer

No padding ('valid') and a stride of 2.

Answer

A 1x1 filter with a stride of 1.

Question 5

What is the key difference between a standard convolution operation and a transposed convolution (often called deconvolution)?

Accepted Answer

Transposed convolution is a mathematical operation that performs learnable up-sampling, increasing the spatial resolution of the feature map.

Answer

Transposed convolution does not use learnable filters.

Answer

Standard convolution is used for up-sampling, while transposed convolution is for down-sampling.

Answer

They are identical operations but are used in different parts of the network.

Question 6

The process in a convolutional layer where a filter slides over the input data, computes element-wise products, and sums them up to create a single value in the output is known as what?

Accepted Answer

Convolution

Answer

Max Pooling

Answer

Activation

Answer

Flattening