WorksheetsWeek4_S1_C2
Total questions: 10
Worksheet time: 5mins
CNNs were originally inspired by the structure of the (a)
Which CNN component is responsible for detecting patterns like edges and shapes?
Fully connected layers
Pooling layers
Convolutional filters
Dropout layers
When a CNN learns to detect a pattern in one location and applies it elsewhere, this is called (a) .
Adding more convolutional layers to a CNN primarily helps the network:
Detect low-level abstract features
Detect higher-level abstract features
Eliminate pooling layers
Replace the need for normalization
Why is zero-padding used in CNNs?
To decrease memory usage
To increase the depth of the network
To preserve the spatial dimensions after convolution
To reduce overfitting
The (a) determines how far the filter moves across the input each step.
The recommended scaling for pixel values in grayscale images is:
-1 to 1
1 to 255
0 to 100
0 to 1
Which of the following statements is NOT TRUE about weight sharing in CNNs?
Each neuron has unique weights
Filters share the same weights across spatial positions
Weights are only shared between different models
It increases the total number of parameters
The main advantage of using convolutional layers is reduced (a) compared to fully connected layers.
The first layer of a CNN usually detects:
Very abstract features
Low-level features such as edges and corners
Complex feature maps
Fully connected vectors
