Deeplearning

Study plan

D-2: making cheat sheet (5H) D-1: making cheat sheet (5H), memorizing equations(3H) D-0: look over (10M)

Professor

Jongseok Lee

Good

Last edited time

Dec 11, 2025 03:08 AM

Cheat sheet

Bad

- Memorizing hw code was absolutely important - Just memorizing equations which is very specific processes in some models was not helpful. I should have focused more on main comcept and stem of subject.

Tendency

# Review It was not that hard. # Problems 1. In object detection, calculate AP given founded bound boxes. 2. Calculate bilinear interpolation and transposed convolution. Can you make the result of transposed conv same as bilinear? Prove mathematically don’t show just examples. 3. Explain the whole processes of self-supervised learning when you use rotation prediction to do pretext task and then use it to classify images. 4. How can you do video classification with 2D CNN and 3d conv? Explain each methods and compare both way’s pros and cons. 5. ⚠️Complete the python code of self attention. (It was hw2)

Term

2-1-Final

Exam note

- After the mid-term - No cheat sheet - No calculator

Exam date

Dec 11, 2025 10:00