CS-601 Machine Learning - Dec 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
   aDefine machine learning and differentiate it from traditional programming approaches. Identify the key components of machine learning?
   How to answer

   • Definition of machine learning as data-driven learning
   • Difference from traditional rule-based programming: data + output vs rules
   • Key components: dataset, features, model, loss function, optimizer, evaluation
2. 7 Marks
   bExplain how a hypothesis function maps input features to output predictions in a machine learning model?
   How to answer

   • Definition of hypothesis function h(x)
   • Mapping from input feature vector to output prediction
   • Role in training and examples for regression/classification
3. 7 Marks
   aDescribe the process of applying data augmentation techniques to expand the size and diversity of a dataset?
   How to answer

   • Purpose of data augmentation to increase size and diversity
   • Common techniques: rotation, flipping, cropping, scaling, color jitter
   • Steps: select transforms, apply, validate to preserve labels
4. 7 Marks
   bDiscuss the importance of the chain rule in the back propagation algorithm for computing gradients efficiently?
   How to answer

   • Chain rule for composite functions in deep networks
   • Backward propagation of error gradients layer by layer
   • Efficiency vs naive gradient computation
5. 7 Marks
   aDescribe the role of the bottleneck layer in auto encoders and how it captures the essential features of the input data?
   How to answer

   • Autoencoder encoder-bottleneck-decoder structure
   • Bottleneck as compressed latent representation
   • How dimensionality reduction captures essential features
6. 7 Marks
   bDiscuss the techniques and challenges of hyper parameter tuning?
   How to answer

   • Hyperparameters vs parameters and tuning objective
   • Techniques: grid search, random search, Bayesian optimization
   • Challenges: computational cost, search space, overfitting
7. 7 Marks
   aExplain the architecture of a Convolutional Neural Networks (CNN) and how it processes input data through multiple layers to extract hierarchical features.
   How to answer

   • CNN layers: convolution, activation, pooling, fully connected
   • Hierarchical feature learning from edges to objects
   • Weight sharing and local connectivity
8. 7 Marks
   bExplain how pooling layers reduce the spatial dimensions of feature maps?
   How to answer

   • Purpose of pooling: downsampling feature maps
   • Max pooling and average pooling operation
   • Benefits: translation invariance, parameter reduction
9. 7 Marks
   aDescribe the role of Tensor Flow in facilitating the implementation and training of CNNs.
   How to answer

   • TensorFlow computational graph and Keras API
   • Support for convolution, GPU training, data pipelines
   • Model building, training and deployment workflow
10. 7 Marks
    bExplain the architecture of an LSTM unit and how it processes sequential data by controlling the flow of information through various gates?
    How to answer

    • LSTM cell structure and gates: forget, input, output
    • Cell state and hidden state update equations
    • How gating controls long-term dependencies in sequences
11. 7 Marks
    aDefine Reinforcement Learning (RL). Explain its elements in detail?
    How to answer

    • Definition of RL as learning by interaction and reward
    • Elements: agent, environment, state, action, reward, policy
    • Return, value function and exploration-exploitation
12. 7 Marks
    bExplain the difference between Value Iteration and Policy Iteration algorithms?
    How to answer

    • Objective of both: optimal policy via Bellman optimality
    • Value iteration: iterative value update then policy extraction
    • Policy iteration: alternating policy evaluation and improvement
13. 7 Marks
    aExplain the fundamental principles of Bayesian learning, including Bayes' theorem and the concept of posterior probability.
    How to answer

    • Bayes theorem: posterior from prior and likelihood
    • Concept of posterior, prior, evidence
    • Bayesian learning as probabilistic updating with data
14. 7 Marks
    bExplain the role of machine learning in computer vision applications.
    How to answer

    • Computer vision tasks: classification, detection, segmentation
    • ML/CNN role in feature learning from images
    • Examples: face recognition, autonomous driving, medical imaging
15. 14 Marks
    Write short notes on any two :

    1. i)Scope and limitations of Machine Learning
    2. ii)Batch normalization
    3. iii)Inception network
    4. iv)Natural language processing
    How to answer

    • Select any two topics with definition and key points
    • Explain working, advantages and limitations
    • Give applications/examples for each chosen topic