H.
NO:----------------------
Code: L22AA R22
J.B. INSTITUTE OF ENGINEERING AND TECHNOLOGY
UGC AUTONOMOUS
II B.Tech. II Semester Regular Examinations, July- 2024
MACHINE LEARNING
[Common to CSM, CSD,AIDS & AIML]
Time: 3 Hours Max. Marks: 60
Instructions: 1. Answer All questions in Part-A.
2. Part-B is “either”, “or” choice and may contain sub-questions.
Q.No Bloom’s
PART-A 10 Marks
Level
1. a. What is overfitting in machine learning? L1 1M
b. What is the purpose of an activation function in a neural network? L2 1M
c. What is the main idea behind bagging? L3 1M
d. What is the main difference between a Markov chain and a Hidden Markov Model? L1 1M
e. What is the Bellman equation in reinforcement learning? L2 1M
2. a. What is the difference between classification and regression? L3 1M
b. What is a hidden layer in a neural network? L1 1M
c. What is stacking in ensemble learning? L2 1M
d. What is the purpose of the transition matrix in an HMM? L3 1M
e. What is the actor-critic method in reinforcement learning? L1 1M
50
PART- B
Marks
3. a. Explain the different types of the machine learning methods? L2 5M
b. Explain the linear regression method with suitable example. L3 5M
[ OR ]
c. Explain the concept of regularization in linear regression. How do Lasso and Ridge L2 10 M
regression address the problem of overfitting, and what are the key differences
between them?
4. a. Explain the Artificial Neural Networks(ANN) with suitable example. L3 5M
b. Discuss the advantages and disadvantages of decision trees compared to other L2 5M
machine learning algorithms, such as logistic regression and support vector
machines (SVMs)
[ OR ]
c. Explain how decision trees work, including how they split data at each node, the L2 10 M
criteria used for splitting and how they handle categorical and numerical data.
5. a. Discuss various features of Boosting. L2 5M
b. Explain features of Bayesian Networks. L3 5M
[ OR ]
c. Disuses various features of multiclass classification L2 5M
d. Explain various features of ensemble learning. L3 5M
6. a. What is a Hidden Markov Model (HMM), and how does it differ from Markov L2 5M
chains?
b. What is BIRCH (Balanced Iterative Reducing and Clustering using Hierarchies), L3 5M
and how does it differ from traditional clustering algorithms like K-means?
[ OR ]
c. Explain the key components of a Hidden Markov Model: states, observations, L3 5M
transition probabilities, and emission probabilities. How are these components used
to model sequential data?
d. Discuss the forward-backward algorithm in Hidden Markov Models. How does it L2 5M
compute the probability of observing a sequence of events given the model?
7. a. Explain the basic components of a reinforcement learning problem: agent, L2 5M
environment, actions, states, rewards, and policies. How do these components
interact in the RL framework?
b. Explain Q-Learning as a model-free reinforcement learning algorithm. How does L3 5M
Q-Learning estimate the quality of actions (Q-values) and update its policy to
achieve optimal behavior?
[ OR ]
c. What are Double Deep Q-Networks (DDQN), and why were they introduced as an L3 10 M
enhancement to traditional Deep Q-Networks (DQN)?
-------
