sp26-cs639 cs639
Practice Midterm 5 Practice Midterm 5
Content locked
The catalog is public. Reading this content requires approved access.
Sign in to readPractice Midterm 5
1. [ML Foundations] Cross-Entropy Loss Computation
2. [ML Foundations] Normal Equations Invertibility
3. [CNNs] Output Dimension with Stride and Padding
4. [RNNs] Vanishing Gradients and LSTM Gates
5. [GNNs] Aggregate-and-Update Message Passing
6. [Self-Supervised Learning] Teacher-Student Update Rule
7. [Scaling Laws] Compute Optimum and Power Laws
8. [Transformers] Attention Head Dimensionality
9. [Transformers] Rotary Positional Encoding (RoPE) vs. Sinusoidal
10. [BERT Architecture] MLM Masking Protocol Details
11. [GPT] From GPT-2 to GPT-3: What Changed
12. [Attention Variants] Complexity Hierarchy
13. [FlashAttention] Tiling and the Softmax Decomposition
14. [Multimodal] ViT Patch Count and Sequence Length
15. [Multimodal: CLIP] Contrastive Loss Symmetry
16. [Multimodal] Flamingo’s Perceiver Resampler
17. [Prompting: ICL] Zero-Shot vs. Few-Shot — When Few-Shot Hurts
18. [Prompting] OPRO — LLM as Optimizer
19. [Cross-Lecture: Foundations + Transformers] Loss Function at Inference Time
20. [Cross-Lecture: SSL + Multimodal] Contrastive Learning in CLIP
21. [Cross-Lecture: Architectures + Attention] Why BERT Cannot Do Zero-Shot CoT
22. [GNN + Self-Supervised] Semi-Supervised Node Classification
23. [Cross-Lecture: Scaling + Prompting] Test-Time Scaling
24. [Cross-Lecture: Multimodal + Prompting] VLM Prompt Engineering
25. [Cross-Lecture: Full Integration] Designing a Retrieval-Augmented Medical VLM