10/14/23, 10:38 AM Medical Image Analysis - - Unit 14 - Week 11
[email protected]
(https://swayam.gov.in)
(https://swayam.gov.in/nc_details/NPTEL)
NPTEL (https://swayam.gov.in/explorer?ncCode=NPTEL) » Medical Image Analysis (course)
If already
registered, click
to check your
Week 11 : Assignment
payment status The due date for submitting this assignment has passed.
Due on 2023-10-11, 23:59 IST.
Course Assignment submitted on 2023-10-10, 23:36 IST
outline
1) Which of the following statements about dropout regularization in deep learning is true? 1 point
How does an Dropout randomly selects neurons to be "dropped out" during training, which helps prevent
NPTEL online overfitting.
course work?
Dropout adds noise to the inputs to help prevent overfitting.
()
Dropout increases the number of neurons in the network to prevent overfitting.
Week 0 () Dropout only works well with small datasets.
Yes, the answer is correct.
Week 1 () Score: 1
Accepted Answers:
Week 2 () Dropout randomly selects neurons to be "dropped out" during training, which helps prevent
overfitting.
Week 3 ()
2) Which of the following is not a common regularization technique used in neural 1 point
Week 4 () networks?
Dropout
Week 5 ()
L1 regularization
Week 6 () L2 regularization
Data augmentation
Y h i
https://onlinecourses.nptel.ac.in/noc23_bt40/unit?unit=47&assessment=146 1/4
10/14/23, 10:38 AM Medical Image Analysis - - Unit 14 - Week 11
Yes, the answer is correct.
Week 7 () Score: 1
Accepted Answers:
Week 8 () Data augmentation
3) Suppose you are working on a binary classification problem with an imbalanced 1 point
Week 9 ()
dataset where the positive class represents only 5% of the data. You train a model and achieve an
accuracy of 95%. Is this a good model?
Week 10 ()
Yes, the model has achieved high accuracy and is therefore good.
Week 11 ()
No, the model may be biased towards the majority class and have poor predictive power for
the positive class.
CNN Training
(unit? It’s hard to say without more information about the problem and evaluation metrics such as
unit=47&lesson precision and recall.
=115)
Both A and C.
Semantic Yes, the answer is correct.
Segmentation Score: 1
(unit? Accepted Answers:
unit=47&lesson No, the model may be biased towards the majority class and have poor predictive power for the
=116) positive class.
Classification
Demo in 4) Consider the algorithm of batch normalization. In the same, the values of and are 1 point
Pytorch (unit? (assume they and the variable hold the same meaning as discussed in the lecture)
unit=47&lesson
=117) Hyperparameters
Quiz: Week 11
Learned Parameters
: Assignment is used to scale the normalised value of (value in a mini-batch)
(assessment?
is used to scale the normalised value of (value in a mini-batch).
name=146)
No, the answer is incorrect.
Week 11 Score: 0
Feedback Form Accepted Answers:
: Medical Image Learned Parameters
Analysis (unit? is used to scale the normalised value of (value in a mini-batch)
unit=47&lesson
=48)
5) Which of the following is true about Grid Search for hyperparameter tuning? 1 point
Week 12 ()
It is a random approach for hyperparameter tuning.
Download It involves systematically iterating over all possible combinations of hyperparameters to find
Videos () the best set.
It is a heuristic approach for hyperparameter tuning.
Transcripts () It involves selecting hyperparameters based on intuition and experience.
Yes, the answer is correct.
Score: 1
Accepted Answers:
It involves systematically iterating over all possible combinations of hyperparameters to find the
best set.
https://onlinecourses.nptel.ac.in/noc23_bt40/unit?unit=47&assessment=146 2/4
10/14/23, 10:38 AM Medical Image Analysis - - Unit 14 - Week 11
6) In which type of neural network architecture are skip connections commonly used? 1 point
Convolutional neural networks
Recurrent neural networks
Autoencoders
Generative adversarial networks
Yes, the answer is correct.
Score: 1
Accepted Answers:
Convolutional neural networks
7) Which of the following is a commonly used CNN architecture for segmentation tasks? 1 point
VGG
ResNet
U-Net
All of the above.
Yes, the answer is correct.
Score: 1
Accepted Answers:
U-Net
8) What are Transposed convolutions? 1 point
Transposed Convolutions are used to upsample the input feature map to a desired output
feature map using some learnable parameters.
Transposed Convolutions are used to downsample the input feature map to a desired output
feature map using some learnable parameters.
Transposed Convolutions are used to upsample the output feature map to a desired output
feature map using some learnable parameters.
Transposed Convolutions are used to downsample the output feature map to a desired
output feature map using some learnable parameters.
No, the answer is incorrect.
Score: 0
Accepted Answers:
Transposed Convolutions are used to upsample the input feature map to a desired output feature
map using some learnable parameters.
9) Which of the following techniques can be used to overcome the class imbalance 1 point
problem in segmentation tasks?
Data augmentation
Weighted loss functions
Oversampling
All of the above
Y h i
https://onlinecourses.nptel.ac.in/noc23_bt40/unit?unit=47&assessment=146 3/4
10/14/23, 10:38 AM Medical Image Analysis - - Unit 14 - Week 11
Yes, the answer is correct.
Score: 1
Accepted Answers:
All of the above
10) What is the purpose of the encoder network in SegNet, a commonly used semantic 1 point
segmentation model?
To compress the input image into a lower-dimensional representation
To expand the lower-dimensional representation into the original input image size
To predict the pixel-wise classification of the input image
To preprocess the input image before passing it to the decoder network
Yes, the answer is correct.
Score: 1
Accepted Answers:
To compress the input image into a lower-dimensional representation
https://onlinecourses.nptel.ac.in/noc23_bt40/unit?unit=47&assessment=146 4/4
