AL-405 Machine Learning - Jun 2024 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 difference between classification and regression models.
   How to answer

   • Define classification (discrete labels) vs regression (continuous values)
   • Compare output type, objective/loss functions
   • Give examples of algorithms for each
   • Mention evaluation metrics for each
2. 7 Marks
   bWhat a hypothesis space represents in machine learning? List some basic characteristics of a hypothesis space?
   How to answer

   • Define hypothesis space as set of all hypotheses learnable by algorithm
   • Explain role in search for best hypothesis
   • List characteristics: size, expressiveness, bias, completeness
   • Mention overfitting vs underfitting relation
3. Marks
   aDiscuss some common limitations of machine learning.
   How to answer

   • Define ML limitations context
   • Discuss data dependency and quality
   • Discuss generalization, bias, interpretability
   • Discuss computational cost and need for expertise
4. 7 Marks
   bHow do neural networks leverage parallel processing to perform computations efficiently?
   How to answer

   • Explain neuron as parallel processing units
   • Describe layers and simultaneous weighted sum + activation
   • Mention GPU/parallel hardware acceleration
   • Explain efficiency for large-scale computation
5. 7 Marks
   aWhat is the Perceptron learning algorithm? Discuss how the Perceptron learning algorithm adjust the weights of connections between neurons to learn from data?
   How to answer

   • Define Perceptron and its learning rule
   • Explain weight initialization and update formula $w = w + \eta (y-\hat{y})x$
   • Describe iterative error correction process
   • Mention convergence for linearly separable data
6. 7 Marks
   bDescribe how neural network architectures are represented graphically or conceptually in machine learning?
   How to answer

   • Describe nodes, layers, weights as graph
   • Explain feedforward and layered conceptual view
   • Mention computational graph and matrix notation
   • Note visualization of architectures
7. 7 Marks
   aExplain the concept of entropy in the context of Decision Trees. How does entropy affect the construction and splitting of nodes in a Decision Tree?
   How to answer

   • Define entropy $H(S) = -\sum p_i \log_2 p_i$ as impurity measure
   • Explain low vs high entropy meaning
   • Explain information gain for split selection
   • Describe recursive node splitting process
8. 7 Marks
   bExplain the difference between linear and logistics regression with example.
   How to answer

   • Define linear regression for continuous output
   • Define logistic regression for classification with sigmoid
   • Compare hypothesis, loss, output range with example
   • Give example datasets for each
9. 7 Marks
   aList the advantages of Support Vector Machine (SVM) and how optimal Hyper plane differ from Hyper plane.
   How to answer

   • List SVM advantages: margin, high-dimension, kernel
   • Define hyperplane vs optimal/maximum-margin hyperplane
   • Explain support vectors role
   • Mention robustness to overfitting
10. 7 Marks
    bDescribe the basic concept of K-means clustering and how it partitions data into K clusters based on similarity.
    How to answer

    • Define K-means clustering objective
    • Explain initialization, assignment, update steps
    • Explain similarity typically Euclidean distance
    • Mention convergence and choice of K
11. 7 Marks
    aHow does adaptive hierarchical clustering adaptively adjust the number of clusters based on the data? Discuss.
    How to answer

    • Define adaptive hierarchical clustering
    • Explain agglomerative/divisive merging criteria
    • Explain data-driven stopping / dendrogram cut
    • Discuss adaptivity to density/shape
12. 7 Marks
    bExplain the difference between the E-step and M-step in the EM algorithm. How does EM handle missing or incomplete data in probabilistic models?
    How to answer

    • Define EM algorithm for latent variables
    • Contrast E-step expectation vs M-step maximization
    • Explain handling of missing/incomplete data via expected likelihood
    • Give example of Gaussian mixture application
13. 7 Marks
    aList the fundamental steps involved in designing and conducting a machine learning experiment.
    How to answer

    • List steps: problem definition, data collection, preprocessing
    • Include model selection, training, evaluation
    • Mention deployment and monitoring
    • Present as ordered experimental workflow
14. 7 Marks
    bWhat is cross-validation in machine learning? Describe its purpose and how it helps in evaluating the performance of machine learning models.
    How to answer

    • Define cross-validation and k-fold concept
    • Explain purpose: generalization estimate, hyperparameter tuning
    • Describe procedure of splits and averaging
    • Explain prevention of overfitting
15. 14 Marks
    Write a short note on any two of the following.

    1. a)Partial Least Squares
    2. b)Activation functions
    3. c)Goals of clustering
    4. d)Measuring classifier performance
    How to answer

    • Select any two topics from list
    • Define each with key formulas/concepts
    • Explain applications and importance
    • Give examples/diagrams where relevant