Deep Learning · TEE · guide

February 2025 re-exam · complete solved paper

1. R25 Q1a · ML versus DL and architecture choice
2. R25 Q1b · Training and validation errors
3. R25 Q1c · CNN hierarchy and stride-two convolution
4. R25 Q1d · BPTT and recurrent strengths
5. R25 Q2a · Activation, ReLU and numeric outputs
6. R25 Q2b · Vanishing/exploding gradients and GRU
7. R25 Q3a · 4–3–2 forward pass and dropout
8. R25 Q3b · Four CNN design explanations
9. R25 Q4a · 64×64 CNN shape and parameter table
10. R25 Q4b · Bipolar AND perceptron
11. R25 Q5a · ResNet shortcut and Inception1×1
12. R25 Q5b · ReLU RNN state with declared orientation
13. R25 Q6a · 3–2–1 forward pass and update w7,w8
14. R25 Q6b · AdaGrad stagnation versus RMSProp
15. R25 Q7a · Long-text sentiment architecture
16. R25 Q7b · Transfer learning
17. R25 Q7c · Learning rate and two losses