AL-405 Machine Learning - Jun 2022 Question Paper

Under each question, How to answer lists the points an answer should cover, outlined by CampusPrep from the syllabus. Check the details against your textbook.

1. 7 Marks
   aExplain the concept of hypothesis space and inductive bias in brief.
   How to answer

   • Define hypothesis space as set of all hypotheses considered by learner
   • Give example e.g. conjunctions / linear functions
   • Define inductive bias as assumptions needed for generalization beyond training data
   • Distinguish restriction vs preference bias with example
2. 7 Marks
   bList and explain perspectives and issues in machine learning.
   How to answer

   • List perspectives e.g. Bayes, PAC, MDL, etc.
   • Explain each perspective briefly
   • List issues e.g. data, representation, overfitting, prior knowledge
   • Explain implications for algorithm design
3. 7 Marks
   aDefine artificial neural network. Explain the biological learning system.
   How to answer

   • Define artificial neural network as interconnected processing units
   • Describe biological neuron: dendrites, soma, axon, synapse
   • Explain biological learning as synaptic strength change
   • Relate ANN model to biological inspiration
4. 7 Marks
   bDescribe the characteristics of back propagation algorithm.
   How to answer

   • Define backpropagation as gradient-based weight update
   • List characteristics: multilayer, differentiable activation, forward-backward passes
   • Mention chain rule and error propagation
   • Mention advantages and limitations e.g. local minima, slow convergence
5. 7 Marks
   aDefine Decision tree learning. List and explain appropriate problems for decision tree learning
   How to answer

   • Define decision tree learning and tree representation
   • Explain nodes, branches, leaves and classification process
   • List appropriate problems: discrete, disjunctive, noisy, etc.
   • Give example domain for each characteristic
6. 7 Marks
   bList and explain the issues in Decision tree learning
   How to answer

   • List issues: overfitting, handling continuous attributes, missing values
   • Explain pruning and Occam's razor
   • Explain attribute selection and computational cost
   • Mention handling noisy data
7. 7 Marks
   aIllustrate K. means clustering algorithm.
   How to answer

   • Define clustering and K-means objective
   • Explain steps: initialize centroids, assign, update
   • Mention convergence criterion
   • Give illustrative example / diagram
8. 7 Marks
   bWrite K means algorithm and separate $\{ 5, 11, 19, 27, 23, 25, 6, 18, 2, 8, 10, 12, 31, 29, 4 \}$ into 3 clusters.
   How to answer

   • Write K-means algorithm steps with distance formula
   • Choose initial centroids and assign points by Euclidean distance
   • Recompute means iteratively till stable
   • Present final 3 clusters for given 1-D data
9. 7 Marks
   aExplain resampling methods in machine learning.
   How to answer

   • Define resampling and need for model evaluation
   • Explain cross-validation, bootstrap, etc.
   • Explain train-validation-test splits
   • Compare methods on bias-variance / cost
10. 7 Marks
    bExplain the concept of measuring the classifier performance.
    How to answer

    • Define classifier performance measures
    • Explain confusion matrix, accuracy, precision, recall
    • Explain ROC, F1-score, error rate
    • Mention need for test set / cross-validation
11. 7 Marks
    aDistinguish between Supervised machine learning and reinforcement learning.
    How to answer

    • Define supervised learning with labeled data and feedback
    • Define reinforcement learning with reward / agent-environment
    • Compare on basis of feedback, goal, data, examples
    • Give applications of each
12. 7 Marks
    bExplain the multi-layer perceptron model with a neat diagram.
    How to answer

    • Define MLP with input, hidden, output layers
    • Draw neat labeled diagram with weights and activations
    • Explain forward propagation and activation functions
    • Mention training by backpropagation
13. 7 Marks
    aExplain parallel processing perception learning in neural networks.
    How to answer

    • Define parallel distributed processing in neural networks
    • Explain perceptron learning rule
    • Explain simultaneous weight updates across units
    • Mention advantages for speed and generalization
14. 7 Marks
    bWhat is the goal of support vector machine (SVM)? How to compute the margin .
    How to answer

    • State goal of SVM: maximum margin hyperplane
    • Define support vectors and margin
    • Give margin formula $2/||w||$ with constraints
    • Explain computation via optimization
15. 14 Marks
    Describe the following:

    1. a)Logistic Regression
    2. b)Issues in machine Learning
    3. c)Requirements of clustering algorithms
    How to answer

    • Define logistic regression with sigmoid hypothesis
    • List issues in ML e.g. overfitting, data quality, interpretability
    • List requirements of clustering e.g. scalability, handling noise, shapes
    • Give brief explanation / example for each