Deeplearning
Deeplearning
Study plan
Professor
Jongseok Lee
Good
Last edited time
Oct 25, 2025 12:30 PM
Cheat sheet
Bad
- Check the calculating and processes again and again until exam ends completely. There was misconception at prob 1
Tendency
# Review
It was quite easy.
# Problems
1. ⚠️ CNN. Stride 3, RGB filter size 5x5, 1 conv layer, 1 MLP. What is the number of weights for 20-class classification?
2. Explain transformer self-attention. Calculate the number of weights. D_k 25, input embedding 10, output embedding 20
3. Write and explain Adam Python code
4. Explain the benefits of depthwise separable convolution with examples
5. Add one residual connection to MLP, compare gradients, and explain why it's used
Term
2-1-Midterm
Exam note
Exam date