Download as DOCX, PDF, TXT or read online from Scribd
Download as docx, pdf, or txt
You are on page 1/ 6
Unit 1: Introduction
Remembering:
1. What is supervised learning?
o A) Learning without labeled data o B) Learning with labeled data o C) Clustering data without supervision o D) Reducing dimensionality of data o Answer: B
Understanding:
2. Which of the following describes a discrete random variable?
A) Takes on any value within a range
B) Takes on a countable number of distinct values C) Follows a continuous probability distribution D) None of the above Answer: B
Applying: 3.
Which rule would you apply to calculate the conditional probability of event A given event B?
A) Bayes' rule B) Independence rule C) Variance rule D) Expectation rule Answer: A
Analyzing: 4. What is the primary difference between independence and conditional
independence?
A) Independence does not depend on another event
B) Conditional independence depends on a third event C) Both A and B D) None of the above Answer: C
Evaluating: 5. Which method is most effective for understanding data distribution?
A) Mean and variance
B) Quantiles C) Probability densities D) All of the above Answer: D Creating: 6. Which approach would you propose to estimate probability densities for continuous random variables?
A) Polynomial curve fitting
B) Quantiles calculation C) Bayesian estimation D) None of the above Answer: C
Unit 2: Linear Models for Regression
Remembering:
1. What does maximum likelihood estimation aim to achieve?
o A) Minimizing the error o B) Finding parameter values that maximize the likelihood o C) Simplifying the model o D) None of the above o Answer: B
Understanding: 2. What is the primary goal of least squares in linear regression?
A) Maximizing the likelihood
B) Minimizing the sum of the squares of the residuals C) Maximizing the sum of the squares of the residuals D) None of the above Answer: B
Applying: 3. Which method would you use to prevent overfitting in linear regression?
A) Maximum likelihood estimation
B) Least squares C) Ridge regression D) Bayesian linear regression Answer: C
Analyzing: 4. What distinguishes Bayesian linear regression from traditional linear
regression?
A) Use of prior distributions
B) Use of least squares C) Only traditional linear regression minimizes residuals D) None of the above Answer: A
Evaluating: 5. Which kernel function would be most appropriate for non-linear data separation?
A) Linear kernel B) Polynomial kernel C) Gaussian kernel D) Both B and C Answer: D
Creating: 6. How would you design a kernel function for a specific problem in a GLM?
A) By using the kernel trick
B) By using a polynomial kernel C) By tailoring the function to the problem's characteristics D) None of the above Answer: C
Unit 3: Mixture Models and EM
Remembering:
1. What is the main purpose of K-means clustering?
o A) To classify data o B) To find clusters in the data o C) To perform regression analysis o D) None of the above o Answer: B
Understanding: 2. What is a mixture of Gaussians?
A) A single Gaussian distribution
B) A combination of multiple Gaussian distributions C) A clustering algorithm D) None of the above Answer: B
Applying: 3. Which method is used in PCA for dimensionality reduction?
A) Singular value decomposition
B) Ridge regression C) Maximum likelihood estimation D) Expectation-Maximization Answer: A
Analyzing: 4. What is the primary goal of the EM algorithm in clustering?
A) To maximize the likelihood of the observed data
B) To minimize the distance between clusters C) To reduce dimensionality D) None of the above Answer: A
Evaluating: 5. Which factor is crucial in determining the number of latent dimensions
in PCA? A) Variance explained by each principal component B) Maximum likelihood C) Dimensionality of the data D) None of the above Answer: A
Creating: 6. How would you design an improved clustering algorithm based on EM
principles?
A) By enhancing initialization techniques
B) By improving convergence criteria C) By incorporating domain-specific knowledge D) All of the above Answer: D
Unit 4: Hidden Markov Models
Remembering:
1. What is a Markov model?
o A) A model with independent states o B) A model where future states depend only on the current state o C) A model with dependent states o D) None of the above o Answer: B
Understanding: 2. What is the Viterbi algorithm used for in HMM?
A) To calculate probabilities B) To find the most probable sequence of states C) To reduce dimensionality D) None of the above Answer: B
Applying: 3. Which algorithm would you apply to calculate the likelihood of a sequence in HMM?
A) Forward algorithm B) Backward algorithm C) Both A and B D) Viterbi algorithm Answer: C
Analyzing: 4. What is the difference between HMM and linear dynamical systems?
A) HMM is used for non-sequential data
B) Linear dynamical systems model continuous data C) HMM models discrete sequences D) Both B and C Answer: D Evaluating: 5. Which factor is crucial for the performance of the Viterbi algorithm in HMM?
A) Initialization B) Convergence C) Scaling factors D) All of the above Answer: D
Creating: 6. How would you propose an improvement to the forward algorithm for better performance in HMM?
A) By optimizing scaling factors
B) By improving state transition probabilities C) By reducing computational complexity D) All of the above Answer: D
Unit 5: Combining Models
Remembering:
1. What is ensemble learning?
o A) Using multiple models to improve prediction o B) Using a single model for prediction o C) Combining data from different sources o D) None of the above o Answer: A
Understanding: 2. What does boosting aim to achieve?
A) Reducing bias B) Reducing variance C) Improving model accuracy D) All of the above Answer: D
Applying: 3. Which method would you use for combining models in ensemble learning?
A) Bagging B) Boosting C) Bayesian model averaging D) All of the above Answer: D
Analyzing: 4. What is the primary goal of adaptive basis function models?
A) To adapt to changes in the data
B) To enhance model flexibility C) To reduce overfitting D) All of the above Answer: D
Evaluating: 5. Which factor is critical for the success of CART in decision-making?
A) Splitting criteria B) Pruning C) Tree depth D) All of the above Answer: D
Creating: 6. How would you design a new ensemble learning method for a specific problem?
A) By selecting diverse models
B) By optimizing model weights C) By incorporating domain-specific knowledge D) All of the above Answer: D
