Scribd
Upload
18 views

Module 5 - QB - Optimization Techniques

Uploaded by

namma.wedding1806
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
18 views

Module 5 - QB - Optimization Techniques

Uploaded by

namma.wedding1806
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

GLOBAL ACADEMY OF TECHNOLOGY

Department of Artificial Intelligence and Data Science


Affiliated to VTU, Accredited by NAAC with 'A' grade
RR Nagar, Bengaluru – 560 098

20ADS63: Optimization Techniques for Data Science

Question Bank for Module 5 – Sub-Gradient Methods and Stochastic


Gradient Descent

Q1. Illustrate the need for subgradients? Explain with an example (6-8 Marks)

Q2. Explain the concept of sub-gradient along with formal representation for
f(x) = |x| (One Dimension) (6-8 Marks)

Q3. Explain the concept of sub-gradient along with formal representation for

(Multi Dimension) (8 Marks)

Q4. Write a Python Code Snippet for computing Gradient using Loss Function.
Also contrast with what changes to be done in code, if computing Sub-Gradient
instead of Gradient. (8-10 Marks)

Q5. Discuss the concept of sub-gradient implementation for Lasso Function.


(4-5 Marks)

Q6. Write a Python code for implementing sub-gradient for Lasso Function.
Give a short explanation for unique concepts used in the code. (8-10 Marks)

Q7. Explain the concept of sub-differential set with the help of an example.
Also list any two properties associated with it. (4-5 Marks)
Q8. List the rules for finding sub-differential set. (4 Marks)

Q9.(i) Explain the concept of Lipschitz Continuity of a function at a point.


(ii) Explain the properties of Lipschitz Continuity.
(iii) When does a function become Lipschitz function? (10 Marks)

Q10. List and explain the three popular types of Gradient Descent Algorithms.
(6-8 Marks)

Q11. Write a Python Code for Implementation for Mini-Batch variant of Stochastic
Gradient Descent. (10-12 Marks)

Q12. Explain the concept of momentum in Stochastic Gradient Descent.


Illustrate the changes in code, for introducing momentum onto mini-batch variant.
(8 Marks)

Q13. Explain the concept of randomizing starting point of Stochastic Gradient


Descent. Illustrate the changes in code, for introducing momentum onto mini-batch
variant. (8 Marks)

Q14. Explain the concept of usage of stochastic gradient descent on Neural


Networks. Illustrate with a Python code the usage of Keras for computing
stochastic gradient descent on Neural Network (8 Marks)

You might also like