Practice Midterm 1

Practice Midterm 1

1. Hypothesis Class and Generalization [ML Foundations: Hypothesis Class]

2. Sigmoid Output Computation [ML Foundations: Logistic Regression]

3. MLP Parameter Count [Deep Learning: MLP Parameters]

4. Convolution Output Dimensions [Deep Learning: CNN]

5. RNN Weight Sharing [Deep Learning: RNN]

6. GCN Receptive Field [Deep Learning: GNN]

7. Self-Supervised Learning: Teacher-Student Methods [Self-Supervised Learning: Pretext Tasks]

8. Triplet Loss Computation [Self-Supervised Learning: Contrastive]

9. Chinchilla Scaling: Compute Allocation [Scaling Laws]

10. Skilllet: Online Curriculum Learning [Scaling Laws: Curriculum]

11. Transformer Architecture: Residual Connections [Transformers: Architecture]

12. Transformer vs. RNN: Parallelization [Transformers: Training]

13. BERT: Random Token Replacement [BERT: Pretraining]

14. BERT: [CLS] Token Representation [BERT: Architecture]

15. GPT: Decoder-Only vs. Encoder-Decoder [GPT: Architecture]

16. GPT: Prefilling vs. Decoding [GPT: Generation]

17. Llama 3.1: Grouped Query Attention [Llama 3.1: Architecture]

18. Linear Attention [Attention Variants: Sparse]

19. FlashAttention: Recomputation Strategy [Attention Variants: FlashAttention]

20. Mamba: Selective State Spaces [SSMs: Mamba]

21. Vision Transformer: Patch Embeddings [Multimodal: ViT]

22. CLIP: Zero-Shot Classification [Multimodal: CLIP]

23. LLaVA Architecture [Multimodal: LLaVA]

24. In-Context Learning: Mechanism [Prompting: ICL]

25. Chain-of-Thought Prompting [Prompting: CoT]