Machine Learning (CS-601) - Important Questions

Explain the definition, scope, limitations and applications of Machine Learning.

Discuss hypothesis functions, hypothesis testing and data distributions.

Explain data preprocessing, data augmentation, normalization and data visualization in machine learning.

Compare traditional machine learning techniques with deep learning for real-world applications.

Explain the different Data Visualization methods in detail.

Explain the role of regression, probability and statistics in Machine Learning.

Explain data preprocessing in detail.

Define machine learning, differentiate it from traditional programming, and identify its key components

List the main types of machine learning algorithms and give example applications for each type.

Explain how a hypothesis function maps input features to output predictions in a machine learning model.

Describe common evaluation metrics used for regression models.

Describe the process of applying data augmentation techniques to expand the size and diversity of a dataset?

Explain multilayer neural networks / multilayer perceptron model in detail, including architecture with diagram, backpropagation, weight initialization, training and testing.

Compare L1 and L2 regularization, and discuss momentum, hyperparameter tuning and autoencoders.

Explain weights, bias, loss function and gradient descent in neural networks.

Explain hyperparameter tuning, its techniques, importance for model performance, and challenges

Define what is meant by gradient descent in machine learning optimization

What are autoencoders and what are their types?

Explain the types of gradient descent

Discuss the sigmoid activation function in detail.

Explain the following concepts in detail. Batch normalization

Discuss the importance of the chain rule in the backpropagation algorithm for efficient gradient computation.

Discuss the advantages and limitations of sigmoid and ReLU activation functions in terms of vanishing gradient problem and output range.

What is the role of the bottleneck layer in autoencoders in capturing essential features of input data?

Explain the architecture and operation of a CNN, including convolution, pooling, dense and loss layers, downsampling, increasing filters, and hierarchical feature extraction.

Explain transfer learning (benefits, transferable features, and applications) and the Inception network architecture in detail.

Explain the process of implementing and training a CNN using TensorFlow and the role of TensorFlow in facilitating it.

Define padding in CNNs and explain the commonly used types of padding and how padding works.

Discuss one-shot learning, dimension reduction and CNN implementation using TensorFlow and Keras.

Discuss padding, stride, flattening, subsampling, input channels, and $1\times 1$ convolution in CNNs.

Define dimensionality reduction and discuss it in detail.

Explain how pooling layers reduce the spatial dimensions of feature maps?

Describe the types of transfer learning, including feature extraction and fine-tuning?

Explain why high-dimensional data is challenging for machine learning algorithms.

Describe how to identify overfitting and underfitting in a CNN model and explain potential solutions.

Explain in detail how principal component analysis (PCA) is carried out for dimensionality reduction.

Compare Value Iteration and Policy Iteration, explaining differences with suitable examples.

Explain the structure/architecture of an LSTM unit and how its gates control information flow to process sequential data.

Explain the concept and working principle of Markov Decision Process (MDP).

Explain the Actor-Critic, Q-Learning and SARSA reinforcement learning algorithms

Explain the Reinforcement Learning framework, Markov Decision Process (MDP) and Bellman equations.

Define reinforcement learning and explain its key elements and concepts in detail.

Explain model-based vs model-free learning and Q-learning.

Explain the Actor-Critic model, including roles of actor and critic, their interaction, and its advantages in reinforcement learning

Define recurrent neural networks and explain the types of RNN architectures.

Describe the significance of n-gram precision and brevity penalty in the BLEU score calculation?

Explain the structural and operational differences between feed-forward networks and recurrent neural networks, and compare vanilla RNNs, LSTM, and GRU.

Explain the Q-learning algorithm assuming deterministic rewards and actions.

Explain the role and applications of machine learning in computer vision.

Explain the applications and utilization of machine learning algorithms in speech processing.

Explain the significance of the ImageNet Competition in the development of deep learning.

Explain Bayes' theorem and the principles of Bayesian learning including posterior probability with an example.

Explain Natural Language Processing.

Write a short note on tokenization.

Describe the concept of support vectors and their role in defining the decision boundary in Support Vector Machines (SVM).

Explain Computer Vision and Reinforcement Learning with appropriate examples.

Use a given Bayesian belief network and dataset to construct conditional probability tables and compute posterior prediction (e.g., car value).

What is Support Vector Machine (SVM)? Explain the concept and role in detail, including application areas.

Define Bayesian learning and explain its impact / role in machine learning.

Write short notes on any two :

Write short notes on any two: