Foundations of Deep Learning — DU B. Voc – Web Designing Previous Year Question Papers

Parts of a question should be attempted together. Q1. i. ii. Explain the role of Deep Leaning in speech and inage recognition. Differentiate between supervised learning and unsupervised learning. iii. Define Dropout and explain how it improves generalization. iv. What is Dropout in Deep Leaning? Explain why it is used. v. vi. Define sequence-to-sequence model with one real-world appfication. What is denoising Auto encoder andwhyis it used? Q2. (a) Explain the sigmoid function and its role in Logistic Regression classification. 8 (b) Compare Momentum, SGD, and Adam optinization methods. Q3. (a) Explain` backpropagation algorithm step by step in neural network training. a) Discuss methods used for avoiding overfitting in Deep Leaning models. Q4. (a) Explain CNN architecture with convolution, activation, and pooling layers. 8 Discuss applications of CNN in computer vision and NLP. 7 (b) Q5. (a) Explain Recurrent Neural Network architecture with diagram. 8 (b) Discuss the vanishing gradient problem and its solutions. 7 Q6. (a) Explain different types of Autuencoders with suitable examples. Explain VAEs and compare them with traditional Autoencoders.' 7 Q7. (a) Explain data preprocessing steps before training a Deep Learning model. 8 (b) Write short notes on Bias-Variance Tradeoff and Multi-task Learning. 7