skills.rest
alirezarezvani
alirezarezvaniCOMMUNITY
Explore all skills by this author →
Website𝕏@RezaRezvaniBlnGitHub

deep-learning-book

Indexes and updates the Goodfellow Deep Learning textbook with chapter guides and training diagnostic tools.

25.3k|3.6k|Updated Oct 19, 2025
Search Tags:#education#regularization#study planner#Machine learning#training diagnostics#Deep Learning#textbook companion
One-click install
npx skills add https://github.com/alirezarezvani/claude-skills --skill deep-learning-book
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: deep-learning-book
Source: https://github.com/alirezarezvani/claude-skills/tree/main/engineering/deep-learning-book/skills/deep-learning-book
Command: npx skills add https://github.com/alirezarezvani/claude-skills --skill deep-learning-book

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Studying the 2016 Deep Learning textbook (Goodfellow, Bengio & Courville) leaves readers unsure which chapters still reflect current practice and how to turn its math into training decisions. This Skill indexes all 20 chapters, flags what has been superseded since 2016 (transformers, AdamW, diffusion, double descent), and converts the book's heuristics into executable diagnostics.

Core Features & Use Cases

  • Chapter and topic navigation: Per-chapter synthesis files, a topic index, glossary, cheatsheet, and a prerequisite-aware reading-path planner script.
  • 2016-to-2026 delta layer: Per-chapter notes on what aged well, what was superseded, and what the book never covered, with confidence levels and citations.
  • Deterministic training tools: Stdlib-only Python scripts for training-failure diagnosis, capacity/regularization planning, and parameter/FLOP/activation-memory calculation.
  • Use Case: A practitioner with val loss far above train loss runs training_diagnostics.py and capacity_planner.py to get an ordered, chapter-cited action plan that accounts for double descent instead of blindly shrinking the model.

Quick Start

Ask the deep-learning-book skill to plan a reading path for learning transformers with an applied background and five hours per week.

Frequently Asked Questions about deep-learning-book

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I plan a reading path through the Deep Learning textbook?▼

Run reading_path_planner.py with a goal, background level, and weekly hours, for example --goal "train a transformer" --background applied --hours-per-week 5. It uses the prerequisite map to order chapters and skips material your goal does not need.

How do I diagnose high training loss or a large train-validation gap?▼

Run training_diagnostics.py with train loss, val loss, gradient norm, and epochs to get a triage verdict, then capacity_planner.py for an ordered action plan. High training error points to capacity or optimization; a large gap points to data or regularization.

Is the 2016 Deep Learning book still worth reading?▼

Yes for the math and diagnosis: chapters 2-5, 11, 16, and 18 hold up well. The delta reference flags what changed, including transformers displacing RNNs, AdamW decoupling weight decay, diffusion descending from score matching, and double descent qualifying the U-curve.

Does this skill reproduce the book's content?▼

No. It is a companion containing only original synthesis, and every chapter file links to the official free chapter at deeplearningbook.org. The rights_and_use.md reference explains why nothing from the copyrighted text is copied.

What topics does the deep-learning-book skill not cover?▼

It excludes reinforcement learning depth, LLM training infrastructure, RLHF and DPO alignment, agentic systems, MLOps tooling, and fairness evaluation, none of which the 2016 book treats. Questions outside the 20 chapters are routed to the delta reference rather than extrapolated.

Do the included Python scripts need external packages?▼

No. All four tools use only the Python standard library, support --help, --sample, and --output json, and return typed exit codes. They run offline with no network calls.

Related Skills

Explore More Education & Research Skills→
duypham2801duypham2801

result-to-claim

Evaluate experimental results against research claims using Codex verdicts.

Community
Advanced
bailynlovebailynlove
1

weights-and-biases

Automates ML experiment tracking, hyperparameter tuning, and model versioning with Weights & Biases.

Community
Advanced
nagisanzeninnagisanzenin
171

data-scientist

Analyzes ML systems usage patterns and designs optimization strategies for reduced costs and improved performance.

Community
Advanced
JesusFigueroa25JesusFigueroa25

ml-engineer

Deploy production machine learning systems with PyTorch, TensorFlow, and model serving.

Community
Advanced