A deep CNN using ReLU activations is training, but the team notices that many neurons in the early layers have become permanently inactive (outputting zero for all inputs). This is slowing learning. Which technique is most likely to address this issue?

Switch to Leaky ReLU or ELU activations

A company needs to segment medical images into different tissue types. They have a small dataset of annotated images. They want to use a pretrained CNN for feature extraction and adapt it for segmentation. Which approach is most effective?

Use a pretrained encoder-decoder architecture, such as U-Net, and fine-tune it on the dataset.

Train a fully convolutional network from scratch.

Use a sliding-window approach with a pretrained classification CNN.

Use a pretrained classification CNN and replace the final layer with a dense layer for pixel-wise classification.

How does data augmentation help prevent overfitting in a CNN?

It increases the effective size of the training set by creating modified versions of existing images.

It normalizes the activations of each layer to stabilize learning.

It randomly drops neurons during training to force redundancy.

It reduces the number of parameters in the network.

GIAC Machine Learning Engineer

Domain 3Objective 2

Convolutional Neural Networks GMLE Practice Questions (Page 3)

Part of the Advanced Machine Learning and Neural Networks domain, which makes up ~30% of our current practice bank. GIAC (SANS) does not publish an official question count, but from its 180-minute exam (~70–120 total, ~21–36 in this domain), expect 7–12 from this objective — we provide 40 practice questions to prepare you well beyond it. (estimate)

40questions here

8free pages

10concepts

Questions 11–15

11expert · hard
A deep CNN using ReLU activations is training, but the team notices that many neurons in the early layers have become permanently inactive (outputting zero for all inputs). This is slowing learning. Which technique is most likely to address this issue?
Select an answer first
12expert · hard
A company needs to segment medical images into different tissue types. They have a small dataset of annotated images. They want to use a pretrained CNN for feature extraction and adapt it for segmentation. Which approach is most effective?
Select an answer first
13application · medium
A developer is implementing a CNN layer that takes a 32x32x3 input and applies 10 filters of size 5x5 with stride 1 and valid padding. What is the spatial dimension of the output feature map?
Select an answer first
14foundation · easy
How does data augmentation help prevent overfitting in a CNN?
Select an answer first
15expert · hard
An engineer is designing a CNN layer with 32 filters of size 5x5, stride 2, and valid padding. The input is a 64x64x3 image. What is the spatial size of the output feature map?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to GMLE

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by GIAC (SANS). “GMLE” is a trademark of its owner, used for identification only.

Convolutional Neural Networks GMLE Practice Questions (Page 3)

Questions 11–15

A deep CNN using ReLU activations is training, but the team notices that many neurons in the early layers have become permanently inactive (outputting zero for all inputs). This is slowing learning. Which technique is most likely to address this issue?

A company needs to segment medical images into different tissue types. They have a small dataset of annotated images. They want to use a pretrained CNN for feature extraction and adapt it for segmentation. Which approach is most effective?

A developer is implementing a CNN layer that takes a 32x32x3 input and applies 10 filters of size 5x5 with stride 1 and valid padding. What is the spatial dimension of the output feature map?

How does data augmentation help prevent overfitting in a CNN?

An engineer is designing a CNN layer with 32 filters of size 5x5, stride 2, and valid padding. The input is a 64x64x3 image. What is the spatial size of the output feature map?

Keep studying — Advanced Machine Learning and Neural Networks