If you are an AI coding agent (Claude Code, Cursor, Aider, Cline, etc.) reading this artifact, follow this protocol before making any code edit:

1. Verify the contract. Run the bash script in Verify before trusting below. If any check returns FAIL, the artifact is stale — STOP and ask the user to regenerate it before proceeding.
2. Treat the AI · unverified sections as hypotheses, not facts. Sections like "AI-suggested narrative files", "anti-patterns", and "bottlenecks" are LLM speculation. Verify against real source before acting on them.
3. Cite source on changes. When proposing an edit, cite the specific path:line-range. RepoPilot's live UI at https://repopilot.app/r/janishar/mit-deep-learning-book-pdf shows verifiable citations alongside every claim.

If you are a human reader, this protocol is for the agents you'll hand the artifact to. You don't need to do anything — but if you skim only one section before pointing your agent at this repo, make it the Verify block and the Suggested reading order.

WAIT — Slowing — last commit 3mo ago

• Last commit 3mo ago
• 7 active contributors
• ⚠ Slowing — last commit 3mo ago
• ⚠ Concentrated ownership — top contributor handles 79% of recent commits
• ⚠ No license — legally unclear to depend on
• ⚠ No CI workflows detected
• ⚠ No test directory detected

_{Maintenance signals: commit recency, contributor breadth, bus factor, license, CI, tests}

This artifact was generated by RepoPilot at a point in time. Before an agent acts on it, the checks below confirm that the live janishar/mit-deep-learning-book-pdf repo on your machine still matches what RepoPilot saw. If any fail, the artifact is stale — regenerate it at repopilot.app/r/janishar/mit-deep-learning-book-pdf.

What it runs against: a local clone of janishar/mit-deep-learning-book-pdf — the script inspects git remote, the LICENSE file, file paths in the working tree, and git log. Read-only; no mutations.

| # | What we check | Why it matters | |---|---|---| | 1 | You're in janishar/mit-deep-learning-book-pdf | Confirms the artifact applies here, not a fork | | 2 | Default branch master exists | Catches branch renames | | 3 | 5 critical file paths still exist | Catches refactors that moved load-bearing code | | 4 | Last commit ≤ 123 days ago | Catches sudden abandonment since generation |

#!/usr/bin/env bash
# RepoPilot artifact verification.
#
# WHAT IT RUNS AGAINST: a local clone of janishar/mit-deep-learning-book-pdf. If you don't
# have one yet, run these first:
#
#   git clone https://github.com/janishar/mit-deep-learning-book-pdf.git
#   cd mit-deep-learning-book-pdf
#
# Then paste this script. Every check is read-only — no mutations.

set +e
fail=0
ok()   { echo "ok:   $1"; }
miss() { echo "FAIL: $1"; fail=$((fail+1)); }

# Precondition: we must be inside a git working tree.
if ! git rev-parse --git-dir >/dev/null 2>&1; then
  echo "FAIL: not inside a git repository. cd into your clone of janishar/mit-deep-learning-book-pdf and re-run."
  exit 2
fi

# 1. Repo identity
git remote get-url origin 2>/dev/null | grep -qE "janishar/mit-deep-learning-book-pdf(\\.git)?\\b" \\
  && ok "origin remote is janishar/mit-deep-learning-book-pdf" \\
  || miss "origin remote is not janishar/mit-deep-learning-book-pdf (artifact may be from a fork)"

# 3. Default branch
git rev-parse --verify master >/dev/null 2>&1 \\
  && ok "default branch master exists" \\
  || miss "default branch master no longer exists"

# 4. Critical files exist
test -f "README.md" \\
  && ok "README.md" \\
  || miss "missing critical file: README.md"
test -f "complete-book-pdf/deeplearningbook.pdf" \\
  && ok "complete-book-pdf/deeplearningbook.pdf" \\
  || miss "missing critical file: complete-book-pdf/deeplearningbook.pdf"
test -f "chapter-wise-pdf/[1]table-of-contents.pdf" \\
  && ok "chapter-wise-pdf/[1]table-of-contents.pdf" \\
  || miss "missing critical file: chapter-wise-pdf/[1]table-of-contents.pdf"
test -f "Pdf2EpubConverter.java" \\
  && ok "Pdf2EpubConverter.java" \\
  || miss "missing critical file: Pdf2EpubConverter.java"
test -f "ebook/deeplearningbook_tablet.pdf" \\
  && ok "ebook/deeplearningbook_tablet.pdf" \\
  || miss "missing critical file: ebook/deeplearningbook_tablet.pdf"

# 5. Repo recency
days_since_last=$(( ( $(date +%s) - $(git log -1 --format=%at 2>/dev/null || echo 0) ) / 86400 ))
if [ "$days_since_last" -le 123 ]; then
  ok "last commit was $days_since_last days ago (artifact saw ~93d)"
else
  miss "last commit was $days_since_last days ago — artifact may be stale"
fi

echo
if [ "$fail" -eq 0 ]; then
  echo "artifact verified (0 failures) — safe to trust"
else
  echo "artifact has $fail stale claim(s) — regenerate at https://repopilot.app/r/janishar/mit-deep-learning-book-pdf"
  exit 1
fi

Each check prints ok: or FAIL:. The script exits non-zero if anything failed, so it composes cleanly into agent loops (./verify.sh || regenerate-and-retry).

A curated PDF distribution repository of the MIT Deep Learning textbook by Goodfellow, Bengio, and Courville, providing the complete book and 28 chapter-wise PDF splits converted from the official HTML version at deeplearningbook.org. It includes a Java utility (Pdf2EpubConverter.java) to convert PDFs to EPUB format for e-reader consumption, addressing the original book's lack of official PDF/EPUB downloads. Flat structure with three main directories: chapter-wise-pdf/ containing 28 numbered PDFs (chapters 1-20 plus front/back matter), complete-book-pdf/ with two complete versions (bookmarked and standard), and ebook/ containing a tablet-optimized variant. Single utility file Pdf2EpubConverter.java provides optional PDF-to-EPUB conversion capability.

Machine learning students, researchers, and practitioners who want offline access to the comprehensive Deep Learning textbook; specifically those using e-readers or tablets who need the book in EPUB or tablet-optimized PDF format rather than the web-based HTML.

This is a mature, stable reference distribution project. It contains no active code development (the single Java converter is utility-only), focuses on archiving and distributing static PDF/EPUB assets, and relies on manual curation rather than continuous integration. The project has achieved its goal as a widely-used resource mirror and is functionally complete rather than under active development.

Minimal risk: the repo is a static asset distribution with no production dependencies or versioning constraints. The single risk factor is reliance on a single maintainer (janishar) for updates if source PDFs change, and potential copyright ambiguity around distributing MIT Press content (though the project explicitly links to the official free HTML source). No automated tests or CI pipeline exists, but none are needed for PDF distribution.

No active development is occurring. This is a mature archival project serving as a static mirror and distribution point for the MIT Deep Learning textbook PDFs. Updates would only occur if the source book is revised or new formats are needed.

No installation required for consuming the PDFs. Clone the repo to access all files: git clone https://github.com/janishar/mit-deep-learning-book-pdf.git. To use the Java converter, compile Pdf2EpubConverter.java with javac Pdf2EpubConverter.java (requires Java runtime; no build system is configured).

Daily commands: No runtime execution needed—this is a download/distribution repo. If using the PDF-to-EPUB converter: javac Pdf2EpubConverter.java && java Pdf2EpubConverter <input.pdf> <output.epub> (exact syntax depends on the converter implementation, which is not shown in the file list).

• README.md — Entry point documenting the repository's purpose as a curated collection of the MIT Deep Learning book in multiple PDF formats.
• complete-book-pdf/deeplearningbook.pdf — The primary bookmarked complete book distribution that contributors should reference as the canonical version.
• chapter-wise-pdf/[1]table-of-contents.pdf — Establishes the organizational structure and chapter numbering convention used throughout the chapter-wise split.
• Pdf2EpubConverter.java — Utility tool for converting PDF formats; essential for understanding any automated processing workflows in the repository.
• ebook/deeplearningbook_tablet.pdf — Tablet-optimized version representing an alternative distribution target that maintainers must keep in sync.

• Complete Book Distributions (PDF, Git LFS (potentially for large files)) — Serve the entire MIT Deep Learning book in multiple quality/format variants
  • Failure mode: Corrupted or mismatched PDF versions; download links broken if files moved
• Chapter-Wise PDFs (PDF segmentation tools, consistent naming conventions) — Provide granular access to individual chapters and part introductions
  • Failure mode: Missing chapters, inconsistent numbering, synchronization drift from complete book
• Pdf2EpubConverter Utility (Java, PDF/EPUB libraries) — Convert between PDF and EPUB formats to support multiple device types
  • Failure mode: Conversion errors, encoding issues, loss of formatting in EPUB output
• README & Documentation (Markdown, GitHub badges/links) — Provide usage guidance, download links, and project context
  • Failure mode: Outdated links, broken badges, unclear instructions confusing users

• Original MIT Press Deep Learning Book → Repository complete-book-pdf/ files — Source material ingestion; one-time or periodic refresh as new editions/corrections released
• complete-book-pdf/ source PDFs → Pdf2EpubConverter.java — Format conversion pipeline for tablet and alternate distribution targets
• Pdf2EpubConverter output → ebook/ and complete-book-bookmarked-pdf/ directories — Storage of converted and optimized versions for distribution
• All PDF variants in repository → GitHub Raw / Release endpoints — Public distribution mechanism; users download via README.md links
• Complete book PDF → chapter-wise-pdf/ directory — Segmentation and extraction of individual chapters for granular access

Add a New Chapter-Wise PDF

1. Create new PDF file following the naming convention [##]<chapter-name>.pdf (chapter-wise-pdf/)
2. Place file in chapter-wise-pdf directory with appropriate numeric prefix matching book structure (chapter-wise-pdf/[##]chapter-name.pdf)
3. Update README.md if adding new content or updating chapter references (README.md)

Generate New Distribution Format

1. Place source PDF in complete-book-pdf directory (complete-book-pdf/source.pdf)
2. Use Pdf2EpubConverter.java to convert format if targeting eBook distribution (Pdf2EpubConverter.java)
3. Store converted file in appropriate directory (ebook/ for tablets, complete-book-bookmarked-pdf/ for bookmarked versions) (ebook/ or complete-book-bookmarked-pdf/)
4. Add download badge and link to README.md (README.md)

Update Book Metadata or Rebuild

1. Review current complete-book-pdf versions to identify which needs refresh (complete-book-pdf/)
2. Use Pdf2EpubConverter to regenerate bookmarked or alternate formats (Pdf2EpubConverter.java)
3. Synchronize chapter-wise PDFs if book structure changed (chapter-wise-pdf/)
4. Update complete-book-bookmarked-pdf with new bookmarked version (complete-book-bookmarked-pdf/deeplearningbook.pdf)

• PDF Format — Universal document format with preserved layout, typography, and cross-platform compatibility for academic text
• Git/GitHub — Version control and distribution infrastructure for managing multiple document variants and accessibility
• Java (Pdf2EpubConverter) — Enables programmatic format conversion to support multiple distribution channels (EPUB for tablets, etc.)

• Maintain multiple PDF copies (bookmarked, standard, tablet-optimized, chapter-wise)

  • Why: Different users have different needs: navigation aids, device-specific optimization, granular chapter access
  • Consequence: Increased repository size (~150-200MB estimated) and maintenance burden to keep versions synchronized

• Store PDFs in Git rather than external storage service

  • Why: Simplicity for end users: direct GitHub download links without extra authentication or API calls
  • Consequence: Repository bloat; scaling issues if frequently re-rendering or updating source materials

• Static document repository with Java converter tool

  • Why: Offline-first approach; users can work with PDFs without network dependency after download
  • Consequence: No dynamic content serving, no personalization, limited real-time updates to book corrections

• Does not provide web-based book reading interface or viewer
• Does not implement DRM or access control; purely open distribution
• Does not offer interactive exercises or code execution environments
• Does not track download analytics or user engagement
• Does not provide the original source materials (LaTeX, InDesign files) for the book

• Duplicate Document Instances (Medium) — complete-book-pdf/, complete-book-bookmarked-pdf/, ebook/: Multiple near-identical copies of the same book stored in different directories and formats, creating synchronization risks and maintenance overhead
• Inactive Java Utility (Low) — Pdf2EpubConverter.java: Java conversion tool present but no evidence of active CI/CD pipeline or automated format regeneration; manual, error-prone process if used at all
• No Version Control of Metadata (Low) — File naming conventions in chapter-wise-pdf/: Chapter numbering embedded in filenames ([##] prefix) rather than in structured metadata; brittle to reorganization and prone to manual errors

• Repository size growth (~150-200MB) (Storage & Bandwidth) — Large binary PDF files in Git history; cloning and pulling the repository becomes slower over time
• Manual synchronization between complete-book and chapter-wise PDFs (Process & Maintenance) — No automated process to ensure chapter-wise files remain in sync when complete book is updated; high risk of stale content

No hidden traps: this is a straightforward static distribution repo. The Java converter file exists but has no visible build configuration (no Maven pom.xml or Gradle build.gradle), so using it may require manually adding a PDF processing library (like Apache PDFBox) to the classpath. The README disclaims that PDFs are derived from the free HTML version and recommends purchasing the official book to support authors.

• Backpropagation & Computational Graphs — Core algorithm explained in Part 2 of the textbook; understanding chain rule and gradient flow is foundational to implementing any deep learning system.
• Regularization (L1/L2, Dropout, Early Stopping) — Essential techniques covered in Chapter 7 to prevent overfitting; directly impacts model generalization in production systems.
• Convolutional Neural Networks (CNNs) — Dedicated chapter (Chapter 9) covering spatial structure exploitation; fundamental architecture for computer vision, which the textbook emphasizes.
• Sequence Modeling & RNNs/LSTMs — Chapter 10 explains recurrent architectures for temporal/sequential data; critical for NLP, time-series, and sequential decision-making tasks.
• PDF-to-EPUB Conversion (via Pdf2EpubConverter) — The repo's utility function; understanding PDF structure and EPUB format standards is needed to adapt or improve the converter for better e-reader rendering.
• Generative Models & Adversarial Networks (GANs) — Part 3 (Chapters 17+) covers advanced generative modeling; essential for understanding modern unsupervised learning and synthetic data generation.
• Optimization Algorithms (SGD, Adam, RMSprop) — Chapter 8 dives into training dynamics; choosing the right optimizer is critical for convergence speed and final model performance.

• fastai/fastai — Companion resource: Fast.ai's deep learning library and courses are built on concepts from this Goodfellow textbook; users often reference both.
• d2l-ai/d2l-en — Alternative deep learning textbook (Dive into Deep Learning) with code examples and interactive notebooks, serving similar readers who want theory + hands-on practice.
• DistilledAI/Energy-Based-Models — Research repo implementing concepts from Part 3 (Deep Learning Research) of the MIT book, showing practical application of theoretical material.
• openai/gpt-2 — Applies deep learning theory from this textbook to language modeling; referenced in the book's external links section.
• tensorflow/tensorflow — The primary framework for implementing all algorithms and models taught in the Deep Learning textbook; essential companion for hands-on learning.

To work on one of these in Claude Code or Cursor, paste: Implement the "<title>" PR idea from CLAUDE.md, working through the checklist as the task list.

Add Java unit tests for Pdf2EpubConverter.java with test fixtures

The repo contains Pdf2EpubConverter.java but has no corresponding test suite. This converter is a critical utility for transforming PDFs to EPUB format. Adding comprehensive unit tests would ensure conversion quality, prevent regressions, and help new contributors understand the conversion pipeline. Test cases should cover edge cases like malformed PDFs, missing metadata, and various PDF structures.

• [ ] Create src/test/java/Pdf2EpubConverterTest.java with JUnit 5 framework
• [ ] Add test fixtures (sample small PDFs) in src/test/resources/
• [ ] Test conversion of chapter PDFs from chapter-wise-pdf/ directory
• [ ] Test batch conversion workflow for multiple chapters
• [ ] Test output validation (check EPUB structure and metadata preservation)
• [ ] Document test execution in README.md

Create automated GitHub Actions workflow to validate PDF integrity and generate checksums

The repo stores 28+ PDF files across multiple directories (complete-book-pdf/, chapter-wise-pdf/, ebook/) but lacks automated validation. A GitHub Actions workflow should verify PDF file integrity, detect corrupted files during commits, generate SHA-256 checksums for all PDFs, and maintain a CHECKSUMS.md file for user verification. This prevents accidental corruption and builds user trust.

• [ ] Create .github/workflows/validate-pdfs.yml workflow file
• [ ] Add validation step to check PDF headers and basic structure integrity
• [ ] Generate and commit checksums to CHECKSUMS.md for all PDFs in complete-book-pdf/, chapter-wise-pdf/, and ebook/ directories
• [ ] Add workflow trigger on PDF file changes in those directories
• [ ] Document checksum verification process in README.md
• [ ] Add badge to README showing last validation status

Add comprehensive chapter-wise extraction documentation and validation utility

The chapter-wise-pdf/ directory contains 28 logically organized files (chapters 1-20, parts, TOC, index, bibliography) but lacks documentation on proper usage, ordering, and how to reassemble them. Create a utility and guide that documents chapter numbering convention, provides a manifest file listing chapter details, and includes a script to validate that all expected chapters are present and correctly ordered.

• [ ] Create CHAPTER_MANIFEST.md documenting all 28 files with chapter numbers, titles, page ranges, and descriptions
• [ ] Create a Java utility ChapterValidator.java to verify all expected chapters exist and have correct naming convention
• [ ] Document in README.md how to use chapter-wise PDFs for selective reading
• [ ] Add a chapters/README.md explaining the [N]chapter-name.pdf naming convention and suggested reading order
• [ ] Create a utility script to reassemble chapters back into complete book format with proper page numbering

• Create a Python conversion script (alternative to Java) that uses pypdf or pdfplumber to convert any chapter PDF to EPUB, stored in a scripts/ directory with usage documentation.
• Add a YAML or JSON metadata file (chapters.yml) cataloging all 28 PDFs with title, page count, publication date, and ISBN, enabling programmatic access and better UI generation (e.g., a web index).
• Build a simple GitHub Pages site (HTML + JavaScript) that lists all available PDFs/EPUBs with search functionality and direct download buttons, improving discoverability beyond the README.

This repository primarily distributes educational PDF content from MIT Press and poses minimal security risks. The main concerns are procedural rather than technical: lack of cryptographic verification mechanisms for downloads, missing security disclosure process, and absence of dependency management documentation. The codebase contains no apparent injection vulnerabilities, hardcoded secrets, or exposed credentials. However, implementers should establish practices for file integrity verification, vulnerability reporting, and dependency scanning to improve the overall security posture.

• Medium · Potential Insecure File Download via Raw GitHub URLs — README.md - Download badge URLs. The README contains direct download links using github.com/raw/master URLs which bypass GitHub's web interface security features. These raw file downloads could potentially be intercepted or manipulated if accessed over insecure connections, and there's no integrity verification mechanism (checksums/signatures) documented. Fix: 1. Document SHA-256 checksums for all PDF files. 2. Use HTTPS exclusively (already done). 3. Consider implementing GPG signatures for file integrity verification. 4. Add checksum validation instructions in README.
• Low · Missing Security Headers Documentation — Repository root. No evidence of security.txt, SECURITY.md, or documented security practices. There is no disclosed process for reporting security vulnerabilities or security update information. Fix: 1. Create a SECURITY.md file with vulnerability disclosure guidelines. 2. Add a .well-known/security.txt file following RFC 9116. 3. Document any security considerations for the repository.
• Low · No Dependency Management or SBOM — Repository - Missing dependency files. The Java file (Pdf2EpubConverter.java) suggests build tools may be in use, but no package manifest files (pom.xml, build.gradle, package.json) are visible in the provided file structure. This makes it impossible to verify if dependencies have known vulnerabilities. Fix: 1. Ensure pom.xml or build.gradle is present in the repository. 2. Use OWASP Dependency-Check or similar tools to scan for vulnerable dependencies. 3. Keep dependencies updated regularly. 4. Publish an SBOM (Software Bill of Materials).
• Low · No Code Signing or Release Verification — Release distribution mechanism. No evidence of signed releases, GPG signatures, or cryptographic verification mechanisms for PDF distributions. Users cannot verify the authenticity of downloaded files. Fix: 1. Implement GPG signing for releases. 2. Publish public key in repository. 3. Document verification steps in README. 4. Use GitHub's code signing features for releases.
• Low · Potential PDF Malware Vector — All PDF files in chapter-wise-pdf, complete-book-pdf, and ebook directories. While PDFs are generally safer than executable files, PDF files can contain embedded malicious content, JavaScript, or exploits. No validation or scanning mechanism is documented. Fix: 1. Verify PDFs with VirusTotal or similar tools periodically. 2. Document source and verification process. 3. Consider adding ClamAV or similar scanning in CI/CD. 4. Clearly label PDFs as from trusted MIT sources.

LLM-derived; treat as a starting point, not a security audit.

• Open issues — current backlog
• Recent PRs — what's actively shipping
• Source on GitHub

Generated by RepoPilot. Verdict based on maintenance signals — see the live page for receipts. Re-run on a new commit to refresh.