π Chandigarh University β MCA/BCA/B.Tech
These Mid-Semester papers are part of the official examinations of Chandigarh University,
designed to test knowledge and practical understanding of core MCA, Computer Applications, and B.Tech subjects.
π‘ Each paper encourages students to think critically, apply concepts, and showcase problem-solving skills β helping them prepare for real-world IT challenges.
Usage Condition:
- π Papers are for reference and study purposes only.
- π§βπ Students should use them responsibly and not for any malpractice.
- π Availability depends on the course and year.
Instructions
-
This question paper consists of three sections. It is compulsory for students to attempt all questions of Section A and Section C.
-
Students need to attempt any one question from question no. 12 and question no. 13 of Section C.
-
Students have to attempt any one question from question no. 11 and question no. 10 of Section C.
Machine Learning
Subject Code: 24CAT-701
Semester: 3 | Time: 3 Hours | Max Marks: 60
Section -A
-
Q. No Statement CO Mapping BT Level 1 State the primary distinction between supervised and unsupervised learning. CO1 2 2 Define the concept of a hyperplane in n-dimensional space. Also determine the advantages of maximizing the margin. CO1 1 3 Explain the mathematical formulation of the Bellman equation in the context of reinforcement learning. CO3 2 4 Discuss the purpose and theoretical relevance of the discount factor (?) in Reinforcement Learning. CO3 1 5 State essential components that contribute to TensorFlowβs scalability and performance. CO4 1
Section B
| Q. No | Statement | CO Mapping | BT Level |
|---|---|---|---|
| 6 | Derive the gradient descent update rule for optimizing linear regression weights. | CO2 | 3 |
| 7 | Analyze how different kernel choices affect SVMβs ability to handle non-linear data. | CO4 | 4 |
| 8 | Analyze how E-greedy policy behaves differently during early and late stages of training in a Q-learning algorithm. | CO4 | 3 |
| 9 | Design and develop a small TensorFlow script that not only constructs a 2-D tensor but also dynamically prints its rank, shape, and data type, demonstrating your understanding of tensor properties. | CO3 | 4 |
Section C-(3*10)
| Q. No | Statement | CO Mapping | BT Level |
|---|---|---|---|
| 10 | Compare supervised learning techniques with unsupervised learning techniques. Provide real-world use cases where each approach is most effective. | CO3 | 5 |
| 11 | Compare and contrast SVM with kernel methods against multilayer perceptron for non-linear classification, also discuss its advantages, limitations, and suitable use cases for each. | CO3 | 5 |
Optional Questions of Section C (Attempt any 1)
| Q. No | Statement | CO Mapping | BT Level |
|---|---|---|---|
| 12 | Differentiate between on-policy and off-policy learning approaches. Analyze how SARSA and Q-learning vary with respect to the source of the next action used in their update rules. | CO5 | 5 |
| OR | |||
| 13 | Explain how to perform linear regression using TensorFlow with data from a CSV file. Describe data preprocessing, model structure, loss function, and optimizer with relevant equations. | CO5 | 5 |
