HIGH-YIELD REVISION GUIDE & CHEAT SHEET
Deep Learning & Neural Networks Com-
plete Semester Guide
High-Yield Comprehensive Guide, Computational Graphs, Receptive Field
Derivations, ResNets & Quick-Cram Appendix
Computer Science & Artificial Intelligence | Course Code: University Of California - Berkeley
- CS 182 / CS 282A | 2024-2025 Comprehensive Edition
📘 WHAT’S INSIDE THIS GUIDE:
✔ Computational Graphs & Backpropagation: Multivariate chain rule, Jacobian matrices, and
gradient flow proofs
✔ Weight Initialization Rigor: Mathematical derivation of Xavier (Glorot) vs. He (Kaiming) variance
preservation
✔ Convolutional Neural Networks: Receptive field propagation formulas and spatial dimension
arithmetic
✔ WORKED STEP-BY-STEP PROBLEM: Exact hand-calculated layer-by-layer receptive field
expansion trace across stacked convolutions
✔ Architectural Innovations: Why ResNet identity skip connections (F(x) + x) guarantee non-
vanishing gradient highways
✔ Optimization Mechanics: Decoupled weight decay in AdamW vs. L2 regularization in Adam and
Cosine Annealing
✔ 10 Deadly Deep Learning Exam Traps (Dying ReLU, test-time BatchNorm running statistics,
1x1 conv channel math)
✔ APPENDIX: 1-Hour Emergency Pre-Exam Quick-Cram Cheat Sheet with master activation table
and 25 active recall flashcards
, Computer Science & Artificial Intelligence · University OfDeep Learning & Neural Networks Complete Semester
California - Berkeley - CS 182 / CS 282A Guide
Disclaimer: This is an independently created academic study aid. It is not affiliated with, endorsed by, or sponsored by any specific
university, textbook author, or publisher.
Independent Study Aid. Not affiliated with or endorsed by any specific university or Page 2 of 7
publisher.
Deep Learning & Neural Networks Com-
plete Semester Guide
High-Yield Comprehensive Guide, Computational Graphs, Receptive Field
Derivations, ResNets & Quick-Cram Appendix
Computer Science & Artificial Intelligence | Course Code: University Of California - Berkeley
- CS 182 / CS 282A | 2024-2025 Comprehensive Edition
📘 WHAT’S INSIDE THIS GUIDE:
✔ Computational Graphs & Backpropagation: Multivariate chain rule, Jacobian matrices, and
gradient flow proofs
✔ Weight Initialization Rigor: Mathematical derivation of Xavier (Glorot) vs. He (Kaiming) variance
preservation
✔ Convolutional Neural Networks: Receptive field propagation formulas and spatial dimension
arithmetic
✔ WORKED STEP-BY-STEP PROBLEM: Exact hand-calculated layer-by-layer receptive field
expansion trace across stacked convolutions
✔ Architectural Innovations: Why ResNet identity skip connections (F(x) + x) guarantee non-
vanishing gradient highways
✔ Optimization Mechanics: Decoupled weight decay in AdamW vs. L2 regularization in Adam and
Cosine Annealing
✔ 10 Deadly Deep Learning Exam Traps (Dying ReLU, test-time BatchNorm running statistics,
1x1 conv channel math)
✔ APPENDIX: 1-Hour Emergency Pre-Exam Quick-Cram Cheat Sheet with master activation table
and 25 active recall flashcards
, Computer Science & Artificial Intelligence · University OfDeep Learning & Neural Networks Complete Semester
California - Berkeley - CS 182 / CS 282A Guide
Disclaimer: This is an independently created academic study aid. It is not affiliated with, endorsed by, or sponsored by any specific
university, textbook author, or publisher.
Independent Study Aid. Not affiliated with or endorsed by any specific university or Page 2 of 7
publisher.
