Neural Network Convolutional Neural Networks 1 β Questions and Answers
Question 1: What is a convolutional layer and what operation does it perform?
- A layer that fully connects all neurons
- A layer that slides learned filters across the input to produce feature maps (Correct answer)
- A layer that reduces dimensions via pooling
- A layer that applies dropout to feature maps
Correct answer: A layer that slides learned filters across the input to produce feature maps
A convolutional layer applies learnable filters (kernels) across the spatial dimensions of the input, producing feature maps that detect local patterns.
Question 2: What does max pooling accomplish in a CNN?
- It learns to detect edges in the input
- It downsamples feature maps by taking the maximum value in each pooling window (Correct answer)
- It normalizes feature map values between 0 and 1
- It applies a learned transformation to reduce dimensions
Correct answer: It downsamples feature maps by taking the maximum value in each pooling window
Max pooling reduces the spatial dimensions of feature maps by selecting the maximum activation within each pooling window, providing spatial invariance.
Question 3: What is the receptive field of a neuron in a CNN?
- The set of neurons it is connected to in the same layer
- The region of the input image that influences that neuron's activation (Correct answer)
- The size of the filter applied to the neuron
- The number of feature maps the neuron belongs to
Correct answer: The region of the input image that influences that neuron's activation
The receptive field is the spatial region of the original input that a given neuron in a deeper layer 'sees' and is influenced by.
Question 4: What is the purpose of padding in convolutional layers?
- To add extra training examples to the dataset
- To preserve spatial dimensions and allow filters to be applied at border regions (Correct answer)
- To prevent overfitting by adding noise
- To increase the number of learned filters
Correct answer: To preserve spatial dimensions and allow filters to be applied at border regions
Padding (typically zero-padding) adds border pixels to the input so that the output feature map maintains the same spatial dimensions and edge information is not lost.
Question 5: What does the stride parameter control in a convolutional layer?
- The number of filters in the layer
- The step size with which the filter slides across the input (Correct answer)
- The depth of the feature maps produced
- The learning rate for that layer's weights
Correct answer: The step size with which the filter slides across the input
Stride controls how many pixels the convolutional filter moves at each step; a stride of 2 halves the spatial dimensions of the output compared to stride 1.
Question 6: What is a 1Γ1 convolution and why is it useful?
- A convolution that applies no transformation
- A convolution that mixes channel information without affecting spatial dimensions (Correct answer)
- A convolution used only for the final classification layer
- A convolution that doubles the spatial dimensions
Correct answer: A convolution that mixes channel information without affecting spatial dimensions
A 1Γ1 convolution performs a linear combination across channels at each spatial position, enabling channel dimensionality reduction or expansion without affecting spatial size.
Neural Network Convolutional Neural Networks 1
What is a convolutional layer and what operation does it perform?