Scribd
Upload
5 views

Unit1_PracticeQuestions

Uploaded by

SINIXTER Op
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
5 views

Unit1_PracticeQuestions

Uploaded by

SINIXTER Op
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

Practice Questions

Unit-1
1. A 3-input neuron has weights 2,5,72, 5, 72,5,7 and inputs 4,3,14, 3, 14,3,1. The transfer function
is linear with a proportionality constant k=2k = 2k=2. Compute the output.
2. A 4-input neuron has weights 1,2,3,41, 2, 3, 41,2,3,4, and the inputs are 5,6,7,85, 6, 7, 85,6,7,8.
Compute the output using a ReLU activation function.
3. Define the mathematical equations for ReLU, Sigmoid, Tanh, and LeakyReLU.
4. Compare the advantages and disadvantages of Sigmoid and ReLU in neural networks.
5. Plot the output of the Tanh activation function for input values ranging from −5-5−5 to 555.
6. Explain the differences between Stochastic Gradient Descent (SGD), Mini-batch Gradient
Descent, and Batch Gradient Descent.
7. Which type of gradient descent is more suitable for large datasets and why?
8. What is the vanishing gradient problem? How does it impact training deep neural networks?
9. Describe two solutions to the vanishing gradient problem with examples.
10. Compare SGD, SGD with Momentum, Adagrad, RMSProp, and Adam in terms of learning rate
adjustments.
11. Describe how momentum helps SGD overcome slow convergence in regions with small
gradients.
12. Implement and compare Adam and RMSProp optimizers for a given dataset.
13. Explain the difference between Kaiming and He initialization techniques. Provide scenarios
where each is preferred.
14. Why is weight initialization important for training deep neural networks?
15. Match the following loss functions to their typical applications:
• Mean Squared Error (MSE)
• Cross Entropy Loss
• Huber Loss
• Mean Absolute Error
16. Explain the advantage of using Huber Loss over MSE in regression tasks with outliers.
17. Differentiate between pretraining and fine-tuning in transfer learning.
18. Give an example of how transfer learning can be applied to medical imaging.
19. What are the symptoms of overfitting and underfitting in a machine learning model? How can
they be mitigated?
20. Explain the trade-off between bias and variance and how it relates to model performance.
21. Define hyperparameters and list three examples in deep learning.
22. For a binary classification problem, create a confusion matrix given the following data:
• True Positives: 50
• True Negatives: 40
• False Positives: 10
• False Negatives: 5
23. Calculate accuracy, precision, recall, and F1-score using the confusion matrix.

You might also like