Advance AI and Deep Learning Question Bank

Unit 1 Advanced AI Basics

1. Compare Classical AI and Modern AI with suitable examples.

2. Explain the role of search methods in Artificial Intelligence. Illustrate with a simple problem.

3. What is reasoning in AI? Explain how reasoning helps an AI system make decisions.

4. Explain the major applications of AI in healthcare, education, agriculture, and finance.

5. Compare rule-based AI systems and learning-based AI systems in terms of development, adaptability, and limitations.

6. Select a real-world problem and explain how AI can solve it. Identify the required data, suitable approach, and expected benefits.

Unit 2 Deep Neural Networks Advanced

1. Explain the basic architecture of ResNet. How do residual connections help train deep neural networks?

2. Explain the basic architecture of DenseNet and compare its connections with those of ResNet.

3. What is overfitting? Explain how Dropout helps reduce overfitting in deep neural networks.

4. Explain Batch Normalization and its effect on training stability and convergence.

5. Explain the working principles of Adam and RMSProp optimizers. Compare their main characteristics.

6. What is hyperparameter tuning? Explain how learning rate, batch size, number of layers, and dropout rate influence model performance.

 

Unit 3 Sequence Models and Attention

1. Explain the working of an RNN and discuss its limitations when processing long sequences.

2. Explain the architecture of an LSTM and the functions of its input, forget, and output gates.

3. Compare RNN, LSTM, and GRU in terms of architecture, handling long-term dependencies, and computational requirements.

4. Explain the basic idea of the attention mechanism with a suitable example.

5. Describe the Transformer architecture at a high level. Compare BERT and GPT in terms of their basic approach and applications.

6. Discuss applications of sequence models, attention, and Transformers in text and image tasks, with suitable examples.

Unit 4 Generative Models

1. Explain the architecture and working of an Autoencoder. Discuss its major applications.

2. Explain the basic concept of a Variational Autoencoder. How does it generate new samples?

3. Compare Autoencoders and Variational Autoencoders in terms of latent representations, training objectives, and generation capabilities.

4. Explain the architecture and working of a GAN, highlighting the roles of the generator and discriminator.

5. Discuss the applications and common training challenges of GANs.

6. Explain an image-generation or text-to-image case study, covering the input, basic workflow, output, applications, and limitations.

Unit 5 Reinforcement Learning and Explainable AI

1. Explain the basic elements of Reinforcement Learning: agent, environment, state, action, reward, and policy, using a suitable example.

2. Explain the basic working of Q-learning. How are Q-values used to select actions?

3. Explain the basic idea of policy gradient methods and compare them with Q-learning.

4. What is Deep Reinforcement Learning? Explain how neural networks are used in it and discuss suitable applications.

5. Explain LIME, SHAP, and Grad-CAM. Compare their purposes and illustrate each with an appropriate example.

6. Discuss bias, fairness, and responsible AI. Suggest measures to develop and evaluate an ethical AI system.