Deep Learning Unit 4 – Optimization & Generalization: AKTU previous year questions
25 AKTU questions from Unit 4 (Optimization & Generalization) asked in 2023–2026, tagged by marks, year and topic. In short: about 34 marks of every paper come from this unit; the most asked topic is Optimization in Deep Learning (10 times); 2 questions came back in a later year. The latest ones are listed below; on the page you can filter them by 2-mark or long questions, topic and repeats.
Unit 4 previous year questions
- Discuss the generalization in machine learning. (2 marks, 2026, Generalization in Neural Networks)
- Explain the optimization problem in deep learning. Why are deep learning loss surfaces highly non-convex? Discuss with suitable diagrams. (7 marks, 2026, Optimization in Deep Learning)
- Compare RNN, LSTM, and GRU networks. Highlight their advantages and use cases. (7 marks, 2026, Recurrent Networks & LSTM)
- What is deep reinforcement learning? Explain its architecture involving policy networks, value networks, and environment interaction. (7 marks, 2026, Deep Reinforcement Learning)
- Define non-convex optimization. (2 marks, 2025, Optimization in Deep Learning)
- Provide examples of optimization algorithms used for deep networks. (2 marks, 2025, Optimization in Deep Learning)
- Discuss the role of non-convex optimization in deep learning. (2 marks, 2025, Optimization in Deep Learning)
- What is generalization in deep learning? (2 marks, 2025, Generalization in Neural Networks)
- How do techniques address the challenges of stochastic optimization in deep learning? (2 marks, 2025, Optimization in Deep Learning)
- Analyze the challenges of stochastic optimization in deep learning. (2 marks, 2025, Optimization in Deep Learning)
- Compare and contrast LSTMs and traditional recurrent neural networks (RNNs). How do LSTMs mitigate the vanishing gradient problem in sequence modeling? (10 marks, 2025, Recurrent Networks & LSTM)
- Explain the role of STNs in addressing spatial invariance in deep learning. (10 marks, 2025, Spatial Transformer Networks)
Most asked Unit 4 topics
- Optimization in Deep Learning – asked 10 times
- Recurrent Networks & LSTM – asked 6 times
- Generalization in Neural Networks – asked 3 times
Unit 4 question pattern
- 2-mark questions: 8 asked in 2023–2026
- 7-mark questions: 3 asked in 2023–2026
- 10-mark questions: 14 asked in 2023–2026
- About 34 marks from Unit 4 in every paper
- 2 questions were asked again in a later year
Unit 4 questions year by year
- 2026: 4 Unit 4 questions asked (23 marks across that year's papers)
- 2025: 16 Unit 4 questions asked (112 marks across that year's papers)
- 2023: 5 Unit 4 questions asked (42 marks across that year's papers)