sp26-cs639 cs639
Practice Midterm 2 Practice Midterm 2
Content locked
The catalog is public. Reading this content requires approved access.
Sign in to readPractice Midterm 2
1. [ML Foundations] Cross-Entropy Loss
2. [ML Foundations] Normal Equations
3. [Deep Learning: CNN] Output Spatial Dimensions
4. [Deep Learning: RNN] Vanishing Gradients and LSTMs
5. [Deep Learning: GNN] GCN Layer Operation
6. [Self-Supervised Learning] Teacher-Student Methods
7. [Self-Supervised Learning: Contrastive] InfoNCE Loss Behavior
8. [Scaling Laws] Compute-Optimal Training
9. [Scaling Laws] Skill-Dependent Scaling
10. [Transformers: Attention] Scaled Dot-Product Computation
11. [Transformers: Architecture] Encoder vs. Decoder Structure
12. [Transformers: Training] Why Causal Masking Matters During Training
13. [BERT] Fine-Tuning for Classification
14. [BERT vs. GPT] Next Sentence Prediction
15. [GPT] Autoregressive Generation Mechanics
16. [Llama] Architecture Innovations
17. [Attention Variants: Linear Attention] Removing Softmax
18. [SSMs: State-Space Models] Discrete Recurrence and Duality
19. [Multimodal: ViT] Patch Sequence Length
20. [Multimodal: CLIP] Zero-Shot Classification
21. [Multimodal: LLaVA vs. Flamingo] Design Comparison
22. [Multimodal: VisualBERT] Cross-Modal Attention
23. [Prompting] Soft vs. Hard Prompts
24. [Prompting: ICL] Recency Bias
25. [Prompting: Chain-of-Thought] Zero-Shot CoT